Introduction

Rice is an important staple crop, accounting for 18.9% of calories consumed globally1. One of the most serious threats for global rice production is rice blast, a disease caused by the filamentous fungal pathogen Magnaporthe oryzae2. Rice blast is responsible for significant crop losses, which if prevented could feed 60 million people each year3. Breeding for rice blast resistance remains the most effective and equitable means to limit rice production losses. The identification and cloning of rice blast resistance genes is an essential first step to deploy knowledge-guided disease resistance breeding strategies4.

Over 100 rice blast resistance loci have been genetically characterized and over 30 rice blast resistance genes have been cloned5. With the exceptions of pi21, encoding a proline-rich metal binding protein6, Pi-d2, encoding a B-lectin receptor kinase7, Ptr, encoding an armadillo repeat containing protein8, and Pi65, encoding a leucine-rich repeat receptor-like kinase9, rice blast resistance genes typically encode intracellular immune receptors of the nucleotide-binding domain leucine-rich repeat (NLR) protein family. Several NLR encoding rice blast resistance genes form allelic series, with various alleles at a given locus having different race-specificity spectra10,11. For example, seven alleles at the Pik resistance locus with varying resistance spectra have been identified12,13. Pi1, Pik, Pik-m, Pik-p, Pik-s, Pik-h and Pike14,15,16,17,18,19. The function of the Pik resistance locus is dependent on a pair of NLRs, one of which contains an integrated heavy-metal associated domain (HMA) responsible for the perception of the M. oryzae effector AVR-Pik14,20. Variation in the HMA domain of Pik is responsible for specific binding to different variants of AVR-Pik, resulting in broad or narrow effector specificity depending on the variant of Pik present21. Like Pik, the Pia locus functions through a pair of NLRs, RGA4 and RGA5, with RGA5 containing an integrated HMA domain22. In contrast to Pik, only two functional (resistance-conferring) Pia alleles have been identified so far. The HMA domains of described RGA5 variants directly interact with two sequence-unrelated M. oryzae effector proteins, AVR-Pia and AVR1-CO3923. While RGA5 acts as a ‘sensor’ NLR, RGA4 is a ‘helper’ NLR that is not involved in effector perception and instead initiates a cell death response after effector recognition by RGA524.

Pi-ta, one of the first NLR resistance genes cloned in rice25, is described to interact directly with AVR-Pita based on yeast two-hybrid and in vitro binding assays26. A single amino acid polymorphism resulting in a serine to alanine substitution at residue 918 has been suggested to differentiate the functional Pi-ta version from non-AVR-Pita recognizing protein variants25. The Pi-ta locus is tightly linked to the rice blast resistance locus Pi-ta2 in the centromeric region of rice chromosome 1227,28. The resistance spectrum of Pi-ta2 was reported to be broader compared to Pi-ta and included all M. oryzae races to which Pi-ta containing rice is resistant, thus making it difficult to differentiate Pi-ta from Pi-ta2 mediated resistance29. The development of near-isogenic lines in the blast-susceptible rice line LTH revealed that the Pi-ta2 resistance spectrum is indeed broader and overlapping Pi-ta30,31. The blast resistance gene Ptr was mapped close to Pi-ta and Pi-ta2 and confirmed to be a specific allele of the armadillo-repeat protein encoded by LOC_Os12g187298. Pi-ta2 was later shown to be allelic and encode a protein identical to the Ptr gene32. Interestingly, mutating Ptr/Pi-ta2 results in susceptibility to AVR-Pita containing isolates, suggesting that Ptr/Pi-ta2 is required for AVR-Pita mediated resistance8,31. To date, a single allele of Ptr has been functionally validated8,31.

Genebanks are an important source of novel resistance genes and allelic variants. However, screening large germplasm collections for agronomically useful traits and identifying the causal genetic loci has traditionally been a laborious and time-consuming process. Digital genebanks like the 3000 rice genomes project33,34 can expedite gene discovery, allowing trait data to be associated to genetic variations at the population level in the form of association studies. Genome-wide-association studies (GWAS) are an effective way to identify genetic loci by leveraging phenotypic and genotypic data from diverse populations without the need to establish bi-parental mapping populations. While several rice blast resistance loci have been identified using GWAS35,36, few studies have validated the underlying genetic cause of resistance, which can improve marker-assisted selection and ensure a unique source of resistance has been identified. Here, we show that GWAS coupled with medium-quality sequencing is an effective means to identify genetic loci, select candidate genes and allow for rapid functional validation using transgenic approaches.

Results

GWAS uncovers peak associations nearby the Pia and Ptr resistance loci

A diversity panel comprising 500 Asian rice (Oryza sativa) accessions was inoculated with six M. oryzae isolates from the Philippines. Isolates from the Philippines have previously been associated with pandemic status as the lineage is distributed globally37. To maximize the likelihood of discovering new resistance genes/alleles, the 500 rice accessions were selected from a larger diversity panel by excluding accessions with some known rice blast resistance genes (see methods). Disease severity was scored 7 days post inoculation (dpi) using a 0-5 scale, with 0 showing no symptoms and 5 showing large eyespot lesions with a diameter greater than 2 mm38. Accessions showing a standard deviation >1.3 across biological replicates were excluded from GWAS, leaving between 390 and 442 accessions with consistent disease severity scores per isolate infection (Supplementary Table 1). The average disease severity scores of the diversity panel followed a bimodal distribution for M. oryzae isolates Mo15-23, Mo15-24, and C923-49, while the disease scores were skewed towards susceptibility for isolates M64-1-3-9-1 and Mo15-125, and towards resistance for isolate IK81-25 (Supplementary Fig. 1).

Considering that the average disease severity scores showed a non-normal distribution (Supplementary Fig. 1), scores were converted to a binary format to allow for binomial treatment of the trait data when performing GWAS. Disease severity scores of <3 were classified as resistant, while disease severity scores ≥3 were considered susceptible. GWAS detected strong associations on rice chromosomes 11 and 12. The chromosome 11 peak was adjacent to the Pia resistance locus. While we detected the association on chromosome 11 for only one isolate (C923-49), similar association profiles were obtained on chromosome 12 for five of the six M. oryzae isolates (Fig. 1a, b, Supplementary Fig. 2). The peak associated SNPs (i.e., the SNPs with the lowest P-values) on chromosome 12 were located between positions 10.79 and 10.92 Mb (based on the Nipponbare reference assembly), close to the Pi-ta and Ptr resistance loci. A second peak on chromosome 12 was detected at position 12.9 Mb for three M. oryzae isolates, IK81-25, M64-1-3-9-1, and Mo15-125 (Supplementary Fig. 2). This association might correspond to a previously described rice blast resistance locus at 13.1 Mb35.

Fig. 1: GWAS detect two strong rice blast resistance associations on chromosomes 11 and 12 near the Pia and Ptr resistance loci, respectively.
figure 1

a Manhattan plots showing associations between individual bi-allelic SNPs and resistance to M. oryzae isolates Mo15-24 (group two) and C923-49 across the twelve rice chromosomes. b Manhattan plots of chromosomes 12 (Mo15-24) and 11 (C923-49) from (a). c Close-up views of genomic regions of co-associated SNPs (haploblock) surrounding the peak SNP. The physical positions of the peak SNPs in the Nipponbare reference assembly for Mo15-24 (10,806,802), Mo15-23 (10,797,768), and C923-49 (6,540,075) are marked by asterisks (Mo15-23 peak SNP location is shown in gray). The linkage blocks surrounding the peak SNPs for Mo15-24 (Chr12: 10,769,576-10,845,095) and C923-49 (Chr11: 6,530,358-6,554,718) are marked by a gray dashed line. Gray scale bars represent 10 kb. Gene identifiers are represented as shortened versions of the Rice Genome Annotation (Osa1) Release 7 by removing LOC_Os from the start of each gene identifier. Exons and introns of known resistance genes are shown as green and white blocks, respectively.

Full size image

Given that similar associations on chromosome 12 were observed for multiple M. oryzae isolates, we sought to determine if these associations suggest a common source of resistance. For all peak SNP positions, a consistent pattern of association to a subset of resistant rice accessions could be observed, and the SNP variants associated with resistance were rarely present in susceptible accessions (<6.1%), except for the M64-1-3-9-1 and Mo15-125 peak associations (Supplementary Table 2, Supplementary Data 1). We refer to the subset of resistant rice accessions carrying peak associated SNPs as ‘associated accessions’. Two distinct groups of rice accessions that carry the chromosome 12 associations between positions 10.79 and 10.92 Mb could be observed, implying two distinct sources of resistance in the rice accessions tested (Supplementary Data 1). One group of indica rice accessions carried SNP variants associated with resistance to M. oryzae isolates IK81-25, M64-1-3-9-1 and Mo15-125 (group one), while the other group of accessions from the indica, aus, and aro subgroups carried SNP variants associated with resistance to M. oryzae isolates Mo15-23 and Mo15-24 (group two) (Supplementary Data 1). The Manhattan plot profile was also suggestive of two distinct groups of associated rice accessions, with the SNP association profiles overlapping between group one isolates, and a different profile observed for group two isolates (Supplementary Fig. 2). Resistance associations corresponding to group one isolates typically carried co-associated SNPs across the 10.79–12.9Mb region implying that this region, which spans the centromere, is tightly linked within group one resistance associated accessions (Supplementary Data 1, Supplementary Fig. 2). For all associations detected in this study, the Manhattan profile and peak associated SNPs was consistent for each isolate infection regardless of gaussian or binomial input data except in the case of Mo15-125 which produced a peak at 10’215’967 and 10’926’845 for binomial and gaussian input data, respectively. All group one and two isolates except for M64-1-3-9-1 are virulent to Pi-ta containing rice (Supplementary Table 3), suggesting that Pi-ta is not responsible for either of the two sets of resistance associations identified. All group one and two isolates except for Mo15-125 are avirulent to Ptr containing rice (Supplementary Table 3), indicating that Ptr may be responsible for the resistance associations to these isolates.

The peak SNP variant associated with resistance to C923-49 on chromosome 11 was enriched in resistant accessions, while being rarely present in susceptible accessions (Supplementary Table 2, Supplementary Data 2). The C923-49 resistance association is adjacent to the Pia locus (Fig. 1c) and the C923-49 isolate is avirulent to Pia (Supplementary Table 3), suggesting that Pia or an allele could be responsible for resistance to C923-49.

To narrow down the genomic regions likely to contain resistance genes on chromosomes 11 and 12, tightly linked SNPs surrounding peak associated SNPs were determined. Linkage blocks surrounding the peak associated SNPs for Mo15-24 (chromosome 12) and C923-49 (chromosome 11) spanned 75.5 kb and 24.4 kb, respectively (Fig. 1c). The resistance gene LOC_Os12g18729 (Ptr) was located within the 75.5 kb linkage block on chromosome 12 (Fig. 1c). The only known functional allele of Ptr was previously shown to confer resistance to the M. oryzae isolate M64-1-3-9-132, but this isolate was virulent on 47 out of 64 rice accessions that carry the group two resistance association (Supplementary Data 1). This indicates that the source of resistance in our study might be a novel allele of Ptr or a closely linked gene.

To account for the possibility that resistant rice accessions contain gene copy-number or presence-absence variation within the defined linkage blocks, we produced de novo assemblies of 10 rice accessions using linked-read sequencing. Three of the 10 accessions carried the group two resistance association (chromosome 12), two carried the group one resistance association (chromosome 12), and a single accession, IRIS_313-12190, carried both chromosome 12 peak SNP associations. Five of the rice accessions selected for sequencing carried the C923-49 resistance association on chromosome 11.

Genome sequencing of resistance-associated rice accessions uncovers functional alleles of Ptr and Pia

Linked read sequencing (10X Genomics) was performed with a coverage of 60-fold or greater per rice accession. The scaffold N50s ranged from 0.78 – 4.6 Mb across the 10 rice accessions, with the total assembly sizes ranging from 411 to 428 Mb, except in the case of IRIS_313-10059 at 502 Mb. BUSCO (Benchmarking Universal Single-Copy Orthologs) scores between 97.7% and 98.4% indicated a high completeness of the 10 assemblies. (Supplementary Table 4).

Assessment of gene presence-absence polymorphisms in regions corresponding to the linkage blocks surrounding peak SNPs revealed no additional genes in the newly sequenced rice accessions compared to the Nipponbare reference, but amino acid (AA) polymorphisms and insertion/deletions in the putative promotor region and introns of Ptr were detected (Supplementary Figs. 3, 4, Supplementary Data 3). When querying the linkage block that partially includes the Pia locus, the five associated accessions carried genomic sequence corresponding to RGA5, while non-associated accessions carried genomic sequence corresponding to LOC_Os11g11810, a sequence divergent NLR present in the reference accession Nipponbare. Given that no additional genes, compared to the Nipponbare reference sequence, were detected within genomic sequence corresponding to the predicted linkage blocks, we decided to assess the sequence diversity of the candidate resistance loci Ptr and Pia. Genomic sequences encompassing the candidate resistance genes were extracted and multiple sequence alignments were performed, and the sequence of proteins encoded by Ptr, and the Pia genes RGA4 and RGA5 were extracted (Supplementary Figs. 3, 4, 6, 7, 8).

The Ptr allele in group two (Mo15-23 and 24) resistance-associated accessions differs from the previously described functional allele of Ptr

Three of the sequenced rice accessions, IRIS_313-10314, IRIS_313-10738, and IRIS_313-8554, which showed the chromosome 12 resistance associations to group two isolates, all contained the same Ptr allele that differs from the previously described functional allele found in the rice cultivar Katy. In comparison to Katy Ptr, the new Ptr variant carried 43 AA substitutions, 14 AA deletions, and 4 AA insertions (Supplementary Figs. 3, 4). A single sequenced rice accession, IRIS_313-12190, carried chromosome 12 associations to both group one and group two isolates. The predicted Ptr protein sequence of IRIS_313-12190 shares homology with the group two resistance-associated accessions, except for four C-terminal AA substitutions, and three N-terminal AA substitutions (Supplementary Figs. 3, 4). Comparing the IRIS_313-12190 genomic sequence to other sequenced accessions suggests that a recombination event occurred somewhere in the genomic region corresponding to Ptr AA positions 136–187, resulting in a Ptr gene that carries a promotor and 5’ genic sequence similar to the allele found in group one associated accessions, with the remaining downstream sequence more similar to group two associated accessions (Supplementary Figs. 3, 4, Supplementary Data 3). This genomic recombination event may account for the presence of both chromosome 12 peak-associated SNPs in IRIS_313-12190, but does not necessarily imply broader resistance specificity.

Rice accessions IRIS_313-10059 and IRIS_313-10879, which carry strong resistance associations to group one isolates, carry Ptr alleles encoding for protein variants that differ from Katy Ptr by six and seven AA, respectively, and also differ from all other sequenced accessions (Supplementary Figs. 3, 4, Supplementary Data 4). In addition to Ptr, IRIS_313-10059 and IRIS_313-10879 were found to carry the reported Pi-ta resistance allele25, which is located ~200 kb away from Ptr at position 10.61 Mb on chromosome 12 (Supplementary Data 1). Unlike the group one resistance-associated accessions, the group two accessions do not carry the Pi-ta resistance allele. With the exception of the Ptr variant found in the group two resistance-associated accessions, the majority of AA polymorphisms differentiating Ptr variants in the 10 genomes we sequenced were located in the C-terminal region of the protein (Supplementary Figs. 3, 4). Considering the large number of AA changes encoded by the Ptr allele found in the group two resistance-associated accessions, we sought to validate the potential resistance function of this highly divergent allele.

Transgenic plants containing the group two (Mo15-23 and 24) resistance-associated allele of Ptr are resistant to rice blast

To test if the group two resistance-associated allele of Ptr is sufficient to confer resistance to rice blast, the susceptible rice accession CO39, which contains the ‘non-functional’ Amane allele of Ptr, was transformed with a construct containing the Ptr allele amplified from the group two resistance associated rice accession IRIS_313-10314. The construct transferred into CO39 contained 10,090 bp of genomic sequence, including 2522 bp of native promotor sequence and 1016 bp of sequence downstream of the stop codon. Four-week-old plants were spray-inoculated with M. oryzae isolates Mo15-23 or Mo15-24. In all lines tested, non-transgenic sibling segregants were susceptible to rice blast infection while heterozygous or homozygous transgenic plants containing the group two Ptr allele, hereby referred to as Ptrb, were resistant (Supplementary Fig. 5).

Ptr is a race-specific resistance gene with different functional alleles

After confirming that resistance is conferred by Ptrb, we tested if the Ptrb resistance spectrum extended to other M. oryzae isolates. Spray inoculations were performed for six isolates with a diverse virulence spectrum (Supplementary Table 3). The Ptrb transgenic lines showed complete resistance to Mo15-23 and Mo15-24 and were susceptible to the other rice blast isolates tested, consistent with a typical qualitative, major resistance gene response (Fig. 2). Ptrb susceptibility to IK81-25 and M64–1–3-9-1 is consistent with the GWAS results and the separation of group one and two resistance associated accessions. The Katy allele of Ptr (Ptra) in the NIL IRBLta2-Re (A NIL in the CO39 background) was previously shown to confer resistance to M. oryzae isolates IK81-25, M101–1–2-9-1 and M64–1–3-9-132, all of which are virulent on transgenic lines carrying the Ptrb allele (Fig. 2). To confirm that Ptra specificity is not the result of a tightly linked gene in the NIL, we performed infection testing with two Ptra mutants32. Both mutants were susceptible to Mo15-23 and M64–1–3-9-1 while their parental line, IR64, was resistant. In summary, both Ptra and Ptrb alleles confer resistance to Mo15-23, while only Ptra is able to confer resistance to M64–1–3-9-1 (Supplementary Table 5). These results confirm that alleles of Ptr confer resistance to different isolates of M. oryzae.

Fig. 2: Ptrb is a race-specific resistance allele.
figure 2

a Multiple sequence alignment of the polymorphic 79-83 C-terminal amino acids of Ptr variants with polymorphic residues relative to the Katy sequence (Ptra) shown as single letter AA code with deletions shown as dashes and identical residues shown as dots. The Mo15-23 and Mo15-24 (group two) resistance-associated rice accessions are shown in green, the IK81-25, M64-1-3-9-1, and Mo15-125 (group one) resistance-associated rice accessions are shown in yellow, and the IRIS_313-12190 accession, which carried resistance associations to both group one and group two isolates at chromosome 12 is shown in gray. A 4 AA deletion that was previously reported to be diagnostic for Ptra is indicated by a blue box. The genotypes presented are classified into five variant subgroups (Ptra – Ptre; indicated in letters a-e to the left of genotype names) based on the allelic diversity observed in 31 genetically diverse genotypes. b Four-week-old Ptrb containing transgenic lines (3 and 16), their non-transgenic siblings (3sib and 16sib), and the susceptible parental line CO39 were spray-inoculated with M. oryzae isolates indicated. Images of representative infected leaves were taken 7 days post-inoculation. Green boxes indicate transgenic plants containing Ptrb. Scale bar = 2 cm. c Infection scoring data from five plants each of 4-week-old Ptrb containing transgenic lines (3 and 16) and their non-transgenic siblings (3sib and 16sib). Scoring data at 7 days post inoculation is represented by box plots where diamonds indicate the average infection score for each genotype. Significant differences between sibling lines infected with each isolate, as determined by a two-sample t-test, are indicated (*** = P < 0.001), n = 5 plants.

Full size image

Comparisons between Ptr amino acid sequence and allele profiling of 3000 rice genomes uncovers significant diversity

After confirming a new, functionally distinct Ptr allele, we sought to compare the sequence variants of Ptr we identified to those present in other sequenced accessions. Comparison of the AA sequence of Ptr from the 10 accessions we sequenced with those of BHA, YT16, Pi1, Pi4, Amane, and Katy8, CO3932, Nipponbare39, O. longistaminata and 12 other diverse rice genomes40 identified 13 different intact protein variants of Ptr (Fig. 3, Supplementary Fig. 4, Supplementary Data 4). We classified these variants into 5 subgroups based on verified resistance function (Ptra and Ptrb), GWAS resistance association (Ptrc), and overall sequence homology (Fig. 3, Supplementary Fig. 4, Supplementary Data 4). The Ptra subgroup is defined by a loss of four AA close to the C-terminus of the protein, a differentiating feature indicated during the first classification of Ptr function8 (Figs. 2a, 3). While all other variant groups are also defined by polymorphisms in the C-terminus of the protein (Figs. 2a, 3), the Ptrb subgroup contains additional AA changes throughout the protein (Supplementary Figs. 3, 4). The variant subgroup c containing the Ptr variants associated with the Ik81-25, M64-1-3-9-1, and Mo15-125 resistance GWAS peak are tentatively named Ptrc (Fig. 3, Supplementary Fig. 4). The two remaining variant subgroups (d and e) contained accessions with similarity to either Nipponbare (subgroup d), or CO39 (subgroup e) (Fig. 3, Supplementary Fig. 4).

Fig. 3: Ptr C-terminal amino acid alignment differentiates Ptr variants into five distinct subgroups.
figure 3

a Amino acid sequence extracted from 30 genomic sequences including; the 10 accessions sequenced here, BHA, YT16, Pi1, Pi4, Amane, and Katy8, CO3932, Nipponbare39, O. longistaminata and 12 other diverse rice genomes40. A full protein sequence alignment can be found in Supplementary Fig. 4. Variant subgroups are labeled a-e with accessions in each subgroup color coded. Accessions containing the Ptr variant encoded by Ptra are colored blue. Accessions with high similarity to the Ptr variants encoded by Ptrb, and the Ptrc resistance association are colored green and yellow, respectively. b Accessions represented in the 3000 rice genomes project were profiled based on non-synonymous SNPs which correspond to amino acid changes differentiating the subgroups in (a). Subgroups can be differentiated by the combination of amino acid changes at alignment positions 879 and 899 and by the characteristic deletion found in Ptra. 259 accessions could not definitively be placed in the different variant subgroups due to missing SNP data or heterozygous base calls.

Full size image

A high density of SNP calls are available in coding region of Ptr for the vast majority of the 3000 rice genomes, thus allowing for allele profiling based on non-synonymous changes that co-associate with group two (Ptrb) resistance-associated accessions, group one resistance associated accessions (Ptrc), or the previously described Katy Ptr allele (Ptra). We further profiled the 3000 rice genomes for non-synonymous SNPs leading to amino acid changes associated with the e and d variant subgroups. Surveying the 3000 rice genome SNP data revealed that 696 of the 3024 rice accessions likely carry Ptrb or a similar variant, 667 accessions carry alleles similar to Ptrc, 66 accessions appear to carry Ptra, and 914 and 440 accessions are similar to the variant d and e subgroups, respectively (Fig. 3b, Supplementary Data 5). This haplotype analysis indicates that Ptra is comparatively rare in the rice gene pool. The allele profiling results we provide (Supplementary Data 5) give breeders and researchers a useful tool to identify, functionally validate, and deploy diverse alleles of Ptr in agronomically relevant accessions.

Rice accessions with the chromosome 11 resistance association at the Pia locus carry variable alleles of Pia resistance genes RGA4 and RGA5

Five of the 10 accessions for which de novo assemblies were obtained carried the chromosome 11 resistance association at the Pia locus. Of these, three carried an RGA4 protein version that was identical to the previously described functional version from the rice cultivar Sasanishiki22, and two carried an RGA4 variant with 66 AA changes relative to Sasanishiki (Supplementary Figs. 6, 7, Supplementary Data 4). This polymorphic variant of RGA4 is more similar to the variant found in the five non-associated accessions than to the variant found in other associated accessions (Supplementary Figs. 6, 7, Supplementary Data 4). A leucine at Sasanishiki AA position 671 of RGA4 is common among all associated accessions and is replaced by a serine in non-associated accessions containing a full-length version of RGA4 (Supplementary Fig. 7).

Sequence homologous to RGA5 was only detected for the five C923-49 resistance-associated accessions (Supplementary Fig. 6). Non-associated accessions instead carried genes homologous to LOC_Os11g11810, which only shares homology to RGA5 in the 5’ genic region encoding the first 148 amino acids. Besides IRIS_313-10059 RGA5, which encodes an RGA5 identical to Sasanishiki, the RGA5 encoded in rice accessions IRIS_313-10314 and IRIS_313-11786 contain 12 AA changes relative to Sasanishiki, including 9 within the HMA domain (Supplementary Figs. 6, 8, 9). The AA sequence of the IRIS_313-10314 and IRIS_313-11786 RGA5 is similar to the accession 93-1141 except for a 3 AA insertion in the HMA domain (Supplementary Figs. 6, 8, 9). IRIS_313-12190 has 54 AA changes relative to the Sasanishiki RGA5, including 13 within the HMA domain (Supplementary Figs. 6, 8, 9). IRIS_313-11360 features a similar AA sequence to IRIS_313-12190 except for a C-terminal frameshift mutation resulting in a premature stop codon and 42 AA missing from the C terminus of RGA5 (Supplementary Figs. 6, 8). This C-terminal frameshift is located immediately adjacent to the HMA domain.

Together, these results indicate that the presence/absence of the ‘sensor’ NLR RGA5 determines rice blast resistance at the Pia locus, and extensive sequence variation exists in both RGA4 and RGA5.

Alleles at the Pia locus encode variable HMA domains of RGA5 and differ from described alleles

The integrated HMA domain of RGA5 binds the M. oryzae effector proteins AVR1-CO39 and AVR-Pia, and is required for resistance to M. oryzae isolates containing these effectors23. Therefore, we compared the amino acid sequences of the HMA domain between the C923-49 resistance-associated accessions and the known functional RGA5 variants from Sasanishiki and CO39. Three additional HMA variants were detected. Variant 3 present in rice accessions IRIS_313-10314 and IRIS_313-11786 differs from Sasanishiki by nine AA, including a three AA insertion (VES) (Fig. 4a). Besides the three AA insertion, variant 3 is otherwise homologous to the HMA domain of the 93-11 allele (variant 2) (Fig. 4a). The Pi60(t) resistance gene from 93-11 was mapped to a region containing RGA4 and RGA5, and 93-11 was also shown to be resistant to AvrPia containing isolates of M. oryzae42. In addition, protoplasts from the accession Peh-kuh-tsao-tu, which contains the same Pia allele as 93-11, were shown to elicit cell death responses when transiently expressing AVR-Pia22. Together these results suggest that the 93-11 allele of RGA5 is functional. The two remaining C923-49 resistance-associated rice accessions, IRIS_313-12190 (variant 5) and IRIS_313-11360 (variant 6) contain RGA5 HMA variants similar to CO39 from which they differ by four and two amino acid changes, respectively (Fig. 4a).

Fig. 4: C923-49 resistance-associated accessions carry amino acid changes in the integrated HMA domain of RGA5 that lead to predicted changes in AVR interaction surface without altering AVR specificity.
figure 4

a Amino acid alignment of the RGA5 HMA domains from known functional (Sasanishiki and CO39), or likely functional (93-11) resistance protein versions, and the C923-49 resistance associated accessions. HMA variants are numbered 1–6 next to genotype name with newly identified functional variants shown in bold. Secondary structural elements are shown beneath the sequence alignment. Conserved residues relative to Sasanishiki are indicated by dots. Accession labels, deletions, and insertions are color coded to match their respective models in (b, c). IRIS_313-11360* contains a frameshift mutation after the HMA domain of RGA5. b, c Views of the interface between HMA•AVR1-CO39 models from opposite sides. Cartoon representation of three HMA•AVR1-CO39 models involving RGA5_Sasanishiki (orange), IRIS_313-11786 (cyan), IRIS_313-12190 (green), and AVR1-CO39 (magenta). Polymorphisms are indicated from the position numbering of the sequence alignment in (a) and are reported in the structural models by black arrows, plain for surface exposed, and dashed for buried or partially buried positions. A26V corresponds to the replacement of A26 of RGA5_Sasanishiki by a valine residue in a buried position. The last E80K polymorphism is not shown in the structure models, which end at V79. The H12 M13 insertion and the G16R or S18C polymorphisms are shown by gray bracket and arrows, respectively. d Infection testing of C923-49 resistance associated accessions with transgenic M. oryzae isolates expressing AVR1-CO39, AVR-Pia or carrying an empty vector (EV). Colored lines either side of accession names indicate the HMA domain sequence aligned in (a) and modeled in (b/c). e Infection scoring data from six or seven 3-week-old plants for each genotype, per infection treatment, except for IRIS_313-10314 (n = 5). Scoring data at 7 days post inoculation is represented by box plots where diamonds indicate the average infection score for each genotype. Significant differences between AVR expressing Guy11 isolates and the empty vector (EV) control as determined by a two-sample t-test, are indicated (*** = P < 0.001, ** = P < 0. 01). Scale bar = 2 cm. Additional representative infected leaves are presented in Supplementary Fig. 10. Infection results are summarized in Supplementary Table 7.

Full size image

To determine if any of the polymorphisms present in the HMA domains of the C923-49 resistance associated accessions are likely to result in changes to the interaction interface between the HMA domain and the detected effectors, we modeled the variable HMA domains from IRIS_313-11786 (variant 3) and IRIS_313-12190 (variant 5), based on the structure of the Sasanishiki HMA domain bound to AVR1-CO3943 (Fig. 4b, c). For the IRIS_313-11786 (variant 3) HMA-AVR1-CO39 complex the model suggested an increase of the binding interface surface area due to additional interacting residues between the N-terminus of AVR1-CO39 and residues in the re-modeled loop between the first helix and the second strand of the HMA domain, but this does not necessarily indicate a stronger resistance response would take place. In accordance with the increased binding interface, the predicted free binding energy of the IRIS_313-11786 HMA-AVR1-CO39 complex is lower (−7.9 ΔiG kcal/mol) compared to the Sasanishiki complex (−4.7 ΔiG kcal/mol) (Supplementary Table 6). Importantly, the re-modeling of this loop is completely compatible with the interface of the complex and does not introduce any clash. Two exposed residues in the variant 3 interaction surface at positions D36Q and E39Q differ from the Sasanishiki HMA domain, but are also present in the likely functional HMA variant from 93 to 11 (Fig. 4) and do not significantly alter the predicted free binding energy of the 93-11 HMA-AVR1-CO39 compared to Sasanishiki (Supplementary Table 6). All remaining amino acid changes in the HMA domain of IRIS_313-11786 are predicted to be buried or are not located at the interaction surface. The IRIS_313-12190 HMA domain (variant 5), which carries a two AA deletion after position 11, also carries two nearby AA substitutions, G16R and S18C. Together with the deletion, these substitutions may influence the structure or positioning of the first alpha helix, which is involved in the HMA-AVR interaction (Fig. 4a, b, c). However, the predicted binding energy of the HMA-effector complex is similar for IRIS_313-12190 (−5.1 ΔiG kcal/mol) and Sasanishiki (−4.7 ΔiG kcal/mol) HMA domains indicating that the perturbations at the complex interface are marginal (Supplementary Table 6).

Infection testing of C923-49 resistance-associated accessions with transgenic M. oryzae isolates indicate that two new Pia alleles also elicit an effector-triggered immune response in the presence of AVR1-CO39 or AVR-Pia

To determine if the C923-49 resistant rice accessions with polymorphic RGA5 HMA domains, perceive AVR1-CO39 and AVR-Pia, infection tests were performed with the M. oryzae isolate Guy11 transformed with an empty vector construct (EV), AVR1-CO3944 or AVR-Pia45. Nipponbare, which does not contain a functional Pia locus, was susceptible to all three isolates of Guy11 (Fig. 4d, e, Supplementary Fig. 10). Aichi Asahi, which contains the Sasanishiki alleles of RGA4 and RGA5, was susceptible to Guy11 EV, but resistant to Guy11 containing AVR1-CO39 or AVR-Pia (Fig. 4d-e, Supplementary Fig. 10). Of the two accessions that have RGA5 HMA domains similar to CO39 (variant 5), IRIS_313-12190 was resistant to AVR1-CO39 and AVR-Pia as indicated by small lesions compared to Guy11 EV infections, while IRIS_313-11360 was susceptible to all Guy11 transformants (Fig. 4d, e, Supplementary Fig. 10). The susceptible reaction of IRIS_313-11360 is likely caused by a frameshift mutation which is present just after the HMA domain and leads to a truncated RGA5 protein (Supplementary Fig. 6, 8), however we cannot discount the possibility that reduced expression or some other mechanism is responsible for susceptibility. IRIS_313-10314 which possesses variant 3 of the HMA domain was resistant to Guy11 EV meaning the response to AVR1-CO39 or AVR-Pia could not be determined (Fig. 4, Supplementary Fig. 10). However, IRIS_313-11786, which also contains HMA variant 3, was susceptible to Guy11 EV and resistant to Guy11 containing AVR1-CO39 or AVR-Pia (Fig. 4d, e). In summary, the HMA domains of both IRIS_313-11786 (variant 3) and IRIS_313-12190 (variant 5) are able to recognize AVR1-CO39 and AVR-Pia. The resistance responses in IRIS_313-12190 also suggests that its divergent form of RGA4, which differs from Sasanishiki RGA4 by 66 AA (Supplementary Fig. 7), is a functional variant. Interestingly, several non-associated accessions carry intact versions of RGA4, which are similar to the functional IRIS_313-12190 RGA4 allele. Comparing RGA4 in non-associated accessions reveals that they encode 6 AA substitutions between alignment positions 55-63 and a single AA substitution at alignment position 671 (Supplementary Fig. 7). Every genotype sequenced regardless of the presence of RGA5, contains an RGA4 allele encoding the auto-active MHD motif variants TYG or MYG (Supplementary Fig. 7), either of which is sufficient for RGA4 mediated cell death24. Interestingly, IRIS_313-12190 features a naturally occurring MYG auto activity variant of the MHD motif which was previously shown to be sufficient for RGA4 cell death activity when transiently expressed in N. benthamiana24.

Given the observed AA changes in the RGA5 HMA domains of both IRIS_313-11786 (variant 3) and IRIS_313-12190 (variant 5), we sought to determine if these changes altered the specificity to M. oryzae isolates IN017 and IN058 which carry the AVR-Pia-H3 allele and are virulent on rice accessions carrying the Pia resistance locus from Sasanishiki or CO3923. IN017 was virulent on IRIS_313-12190 and IN058 was virulent on IRIS_313-11786 indicating that AVR-Pia-H3 does not elicit an effector-triggered immune response in either accession (Supplementary Fig. 11). When independently testing the influence of the R43G and F24S AA polymorphisms which differentiate AVR-Pia-H3 from Avr-Pia, IRIS_313-12190 (variant 5) was susceptible to both mutant versions of AVR-Pia (Fig. 5). However, IRIS_313-11786 (variant 3) showed a similar level of resistance to Guy11 expressing AVR-Pia (R43G) as to Guy11 expressing AVR-Pia, but was susceptible to Guy11 expressing AVR-Pia (F24S) (Fig. 5). These results are interesting in the context of plant-pathogen co-evolution; where strong selective pressure is placed on plant-perceived effectors like AVR-Pia. While the complete loss of dispensable effectors is common, effector mutations represent an alternative path for pathogens to evade effector-triggered immune responses. These results imply a possible evolutionary path for the selection of single AA changes in AVR-Pia whereby (R43G) could give rise to improved pathogen performance until encountering RGA5 HMA variant 3 which would strongly select for the F24S mutation. This hypothesis is further supported by the observed partial reduction in resistance to Guy11 expressing AVR-Pia (R43G) in CO39 (variant 4) (Fig. 5), and Kitaake45 (variant 1) which would also select for to complete evasion conferred by the further addition of the F24S mutation.

Fig. 5: HMA domain variants respond differently to AVR-Pia amino acid changes which give rise to the virulent AVR-Pia-H3.
figure 5

a Infection testing of RGA5 HMA domain variant containing accessions with transgenic M. oryzae isolate Guy-11 expressing AVR-Pia, AVR-Pia (R43G), AVR-Pia (F24S) or carrying an empty vector (EV). Scale bar = 2 cm. b Infection scoring data from eight 3-week-old plants for each genotype, per infection treatment, except for IRIS_313-11786 infected with Guy11-EV (n = 6). Scoring data at 7 days post inoculation is represented by box plots diamonds indicate the average infection score for each genotype. Groups which do not differ significantly from each other based on a 95% confidence interval are labeled with the same letter (one-way ANOVA, Tukey’s HSD). Colored lines either side of accession names indicate matching HMA domain sequence aligned in Fig. 4a.

Full size image

Pia appears most commonly in indica rice accessions

Identification of Pia haplotypes in the 3000 rice genomes project33 was limited by the fact that homology to the reference Nipponbare genome, on which the SNP data is based, is only maintained in the first 148 AA of RGA5. Still, it was possible to assign 9 associated non-synonymous polymorphisms to 1050 accessions that are likely to encode RGA5 rather than the sequence divergent LOC_Os11g11810 found in Nipponbare (Supplementary Data 2). Comparing the two new functional HMA domain variants of RGA5 we identified to available de-novo genome assemblies shows that these variants are not restricted to one subgroup of rice. The variant 5 form was detected in both indica and japonica rice, while variant 3 was only detected in indica rice accessions (Supplementary Data 4). All of the accessions we compared that did not contain RGA5 were japonica rice accessions. Indeed, SNP profiling across RGA5/LOC_Os11g11810 for the 3000 rice genomes suggests that RGA5 appears more commonly in indica rice than in japonica. In total, 966 of the 1254 accessions that show an RGA5-like haplotype are indica accessions, whereas only 823 of the remaining 1770 rice accessions that do not show an RGA5 haplotype were indica accessions. The higher prevalence of the RGA5 haplotype in indica rice may suggest an introduction early in the domestication of indica rice, consistent with the single origin of indica rice from the intercrossing of japonica rice and a wild rice relative46.

Discussion

In this study, we present a rapid strategy to identify and validate rice blast resistance genes following GWAS. Key to this achievement was the generation of medium-quality assemblies of resistant rice accessions from our diversity panel that carried peak associated SNPs. Disease resistance genes are often located in highly dynamic regions of the genome, where gene copy number and presence-absence variations are frequent. Relying on reference assemblies that do not contain the gene of interest bears the risk of missing new functional alleles47,48. The critical step of generating assemblies from the GWAS panel maximized the likelihood of identifying functional alleles for validation. Performing GWAS using sequenced and well-maintained populations, like the 3000 rice genomes project, allows for profiling of all accessions based on the presence of SNPs likely to be associated with particular resistance alleles. This greatly enhances the capability of breeding programs and prevents investment into the breeding, mapping or validation of resistance genes that have already been described.

We identified a resistance-conferring Ptr allele, designated Ptrb, indicating that Ptr forms an allelic series with multiple functional alleles. Allelic series with multiple functional alleles exhibiting various recognition spectra to various pathogen isolates are common for NLR-encoding genes13,49. However, few allelic series for non-NLR encoding genes have been described. In maize, various alleles of a maize wall-associated kinase gene confer resistance against the fungal Northern corn leaf blight disease50. The wheat powdery mildew resistance gene Pm4, encoding an atypical kinase-MCTP protein, has two known functional alleles51. Ptr represents an additional example of an allelic series for an atypical disease resistance gene. The discovery of Ptrb in this study adds valuable biological insight into the possible function of this atypical disease resistance gene, indicating that Ptr might be involved in the direct or indirect perception of M. oryzae effectors consistent with its qualitative resistance response and race-specificity.

In addition to Ptrb, we identified an additional Ptr allele, tentatively designated Ptrc, that was found to be tightly linked to an IK81-25, M64-1-3-9-1 and Mo15-125 resistance association. Mo15-125 was previously shown to be virulent on Ptra containing rice32, but Mo15-125 was avirulent on the majority of rice accessions carrying the IK81-25, M64-1-3-9-1 and Mo15-125 resistance association at the Ptrc locus. This suggests that Ptrc could confer resistance to M. oryzae isolates that are virulent on both Ptra and Ptrb. In support of the conclusion that other Ptr alleles might be novel sources of rice blast resistance, a rice blast resistance QTL from the weedy rice ecotype BHA which confers a resistance distinct from Pi-ta and Ptr/Pi-ta-2 (Oryza spp.) was mapped to a 186,137 bp region on chromosome 12 that spans Ptr52. Recently, a unique Ptr allele was shown to be present in the weedy rice donor “BHA” suggesting that this allele of Ptr or a nearby gene is responsible for conferring blast resistance53. BHA Ptr encodes a similar protein to subgroup d accessions like Nipponbare, but with AA changes in the C-terminal region that could imply unique specificity (Fig. 3, Supplementary Fig. 4). The Pi57(t) rice blast resistance gene introgressed from the wild rice Oryza longistaminata was mapped to a 51.7 kb region which includes Ptr54. The available genomic sequence of O. longistaminata shows a Ptr allele encoding a protein sequence more similar to Ptrb than any of the other variants we observed, however, two AA polymorphisms are present in the C terminal region relative to Ptrb (Fig. 3, Supplementary Fig. 4, Supplementary Data 4). Given that the mapped region containing Pi57(t) contains just 6 genes54, there is a reasonable likelihood that a functional allele of Ptr is responsible for the Pi57(t) locus derived from O. longistaminata.

Unlike Ptrb and Ptrc, the two additional functional alleles of the Pia resistance genes RGA4 and RGA5 we identified, did not show differential resistance responses to M. oryzae isolates expressing either AVR-Pia or AVR1-CO39. This is despite these Pia alleles encoding for AA changes in the HMA effector binding domain of RGA5 (Supplementary Figs. 8, 9), or in the case of one genotype (IRIS_313-12190), 66 AA changes in the helper NLR RGA4 (Supplementary Fig. 7). Interestingly, several non-associated accessions that do not carry RGA5 had similar intact alleles of RGA4 encoding only 8 AA changes relative to the newly identified functional RGA4 (IRIS_313-12190) (Supplementary Fig. 7). This diversity in RGA4 is interesting, given that RGA4 is an auto-active NLR, triggering resistance responses in the absence of RGA524. All intact variants of RGA4 we identified feature the critical auto-active variants of the MHD motif (TYG or MYG) required for RGA4 mediated resistance responses24 (Supplementary Fig. 7). A recently cloned pair of NLRs underpinning the Pias resistance locus are syntenic to RGA4 and RGA5, with Pias-1 sharing homology to the divergent form of RGA4 we identified in IRIS_313-12190 (Supplementary Fig. 7), and Pias-2 sharing homology to the Nipponbare gene LOC_Os11g1181055 (Supplementary Fig. 12). Based on transient expression assays, RGA4 and Pias-1 induced cell-death is suppressed by RGA5, except when AVR-Pia is also present55. This result supports our observation that RGA5 can function in the presence of divergent forms of RGA4/Pias-1, as was the case for IRIS_313-12190 (Fig. 4, Supplementary Fig. 7). When querying available genomic sequences, we identified 12 genomes which contain homologs of Pias-2/LOC_Os11g11810 rather than RGA5 (Supplementary Fig. 12, Supplementary Data 4). Four of the genomes contained Pias-2-like (>99.80% sequence identity) alleles which encode one to three AA changes relative to Pias-2, while all remaining genomes encode LOC_Os11g11810-like variants that do not contain the 91 C-terminal AA of Pias-2 and instead carry a shorter C-terminus (Supplementary Fig. 12). All of the Pias-2-like alleles encode the same C-terminal AA change H1287Y when compared to Pias-2 (Supplementary Fig. 12).

The N-terminal portion of the coiled-coil domains of RGA41–130 and RGA51–133 were shown to interact in Y2H assays, but the CC domain of RGA51–173 alone was insufficient to suppress RGA4 auto-activity24. With the exception of 9 AA changes, protein homology for the N-terminal 148 AA is maintained between RGA5 and Pias-2/LOC_Os11g11810. Besides these 148 AA, Pias-2/LOC_Os11g11810 is otherwise divergent from RGA5. Given that Pias-2 is able to suppress RGA4 auto-activity, the divergent C-terminus encoded by LOC_Os11g11810-like alleles could represent an additional receptor component that releases RGA4/Pias-1 to trigger a resistance response similar to the Pia and Pias NLR pairs.

This study provides a robust conceptual approach for rapid gene cloning and assessment of genetic diversity, which has led to new biological insights into rice blast resistance. These results add to improving knowledge-guided disease resistance breeding strategies, for example through the deployment of optimal disease resistance gene stacks.

Materials and Methods

Selection of rice blast diversity panel and infection testing of rice accessions

Accessions from the 3000 rice genomes project were selected based on the absence of some known rice blast resistance genes. This was executed by using gene specific markers and bioinformatic analyses to screen for genotypes containing Pi2/9, Pik, and Pi3/5. Accessions were further selected based on their country of origin and genotype subgrouping, resulting in a set of 500 genetically diverse accessions for infection testing (Supplementary Data 1). Infection testing was performed using a set of diverse M. oryzae isolates from the Philippines which was shown to exhibit varying virulence spectra when infecting near-isogenic rice lines containing different rice blast resistance loci (Supplementary Table 3). For each M. oryzae isolate infection, three replicates of 10 plants from each of the 500 rice accessions were grown for 14 days and then spray-inoculated with M. oryzae spore solutions (1–2 × 106 spores/ml). Inoculated plantlets were covered and incubated in cooled conditions (max 28°C) overnight, then transferred and kept in a misted greenhouse for 6 days. Disease assessment was performed 7 days post inoculation. Lesion type for each plant (maximum 10 plants per accession, per isolate infection) was recorded based on the evaluation system published by JIRCAS38. A summary of infection testing data is included in Supplementary Data 1.

Genome-wide-association-studies and investigating candidate gene regions

GWAS was performed using the “3K RG 1M GWAS SNP Dataset, all chromosomes” from the Rice SNP-seek database (https://snp-seek.irri.org/_download.zul;jsessionid=CF24BFBA8A7CEEAD273D6EAC4467EC98). BED/FAM/BIM files were converted to TFAM/TPED files using PLINK56. GWAS were performed using the R package GenABEL57. To correct for population structure, a linear mixed (polygenic) model was estimated using the polygenic function58 and association was determined using the mmscore function59. Both binomial and Gaussian treatments of infection data were used as inputs for GWAS. Infection data was cleaned prior to analysis by removing accessions with a spread of infection scores across 5 or 6 units on the 0–5 infection scale, or with a SD > 1.3. Using the same SNP set as used for GWAS, candidate gene regions surrounding peak associated SNPs were defined using the Solid Spine of LD function in Haploview60. SNPs in associated and non-associated accessions were extracted using the Rice SNP-seek database61. The SNP variants for each accession, at each peak associated SNP, were tabulated alongside resistance scores to determine which SNP variants were commonly associated with rice blast resistance to a particular M. oryzae isolate.

Genome sequencing and de novo assembly

High molecular weight DNA extraction from 2-week-old rice leaves was performed62. Genomic DNA was confirmed to be larger than 100 kb on average using an Agilent Femto Pulse System. DNA integrity was confirmed by agarose gel electrophoresis and concentration was determined using a Life Technologies Qubit Fluorometer. Genome sequencing was performed using linked read sequencing63. Library preparation was performed using a 10X genomics, Chromium™ Genome Library, Gel Bead and Multiplex Kit according to manufacturer’s instructions. 150 bp paired-end sequencing was obtained using an Illumina HiSeq X Ten system. Genome assembly was performed using the 10X Genomics Supernova - 2.1.1 software package for de novo assembly64. The de novo assembled genomes were queried by BLAST search to identify scaffolds containing strong homology to candidate resistance genes. The rice accessions carrying the Mo15-23 and Mo15-24 resistance association on chromosome 12 all carried varying sizes of missing sequence 2 kb upstream of the Ptr start codon. This region was separated by an N string during scaffold assembly due to linked reads being assigned to the same scaffold, but short reads not being able to be assembled across the sequence gap. Primers (Supplementary Data 6) were designed to span this region and resolve the complete genomic sequence for the accession IRIS_313-10314 revealing a duplicate repeat of 108 bp.

Ptr b rice transformation and infection testing

The genomic sequence of Ptrb was amplified from IRIS_313-10314 genomic DNA in two fragments with primers at the distal 5’ and 3’ ends of the Ptrb genomic region containing a 15 bp overlap to linearized (digested with HindIII and BsrGI) pIPKB001 binary vector65. Internal primers were positioned such that an 18 bp overlap between the two genomic fragments would be generated (Supplementary Data 6). Genomic fragments were fused with the linearized pIPKB001 binary vector by In-Fusion cloning (Takara Bio Inc). Correct assembly of the intact Ptr sequence in pIPKB001 was confirmed by Sanger sequencing. Transformation of CO39 and isolation of homozygous single copy transformants and sibling lines was performed66. Four-week-old plants were spray-infected with a solution containing 40,000 M. oryzae spores/ml in water containing 0.1% Tween®20. Following infection, plants were placed in trays containing water, covered with a wet plastic hood, and left in the dark for 16 h. Hoods were removed from the plants and they were left to dry before transferring into 12-h photoperiod, 28 °C day and 23 °C night temperature. Symptoms were analyzed seven days after inoculation, on the youngest fully expanded leaf at the time of inoculation. Representative infection scoring leaf images and associated classifications can be found in Supplementary Fig. 13.

Structural modeling of Pia HMA domain variants

The RGA5 HMA domain variants were modeled on the SWISS-MODEL web site with the HMA domain of the Sasanishiki RGA5 complexed with AVR1-CO39 (PDB: 5ZNG) as a template. For IRIS_313-11786, the ASVNGVESMQ-loop sequence was modeled by RCD+: Fast loop modeling server67. The loop with the least clashes with AVR1-CO39 in the complex was selected. This HMA AVR1-CO39 complex was further refined on the GalaxyWEB server68 and the complex model with the lowest energy was selected.

Infection testing of Pia variants using Guy11 transformants

To produce spores, M. oryzae strains were grown for 5 days at 25 °C on rice flour agar (Ref. 69). Infection assays were performed by spraying 3-week-old rice plants with a suspension of M. oryzae conidiospores adjusted to 35,000 spores/ml in water containing 0.5% of gelatin, incubating sprayed plants at 100% humidity in the dark at 23 °C for 16 h, then transferring them into a growth chamber with 12-h photoperiod, 28 °C day and 23 °C night temperature69. Symptoms were analyzed 7 days after inoculation on the youngest leaf fully expanded at the time of inoculation. Representative infection scoring leaf images and associated classifications can be found in Supplementary Fig. 13.

Statistics and reproducibility

Replicates (n) are defined as individual plants or a single sample from individual plants. For infection studies, a single inoculated leaf of the same age from each plant is assessed and this constitutes a biological replicate. Sample sizes and statistical tests are indicated in figure legends.