Resistance to AVR-Rmg8 maps to a 5.3 Mbp region on chromosome 2A of wheat
We selected two MoT isolates to phenotype resistance to AVR-Rmg8. The first isolate (Py 15.1.018) carries the eI allele of AVR-Rmg8 and is virulent against cultivar Jagger and the CBFusarium ENT014 wheat line. (Supplementary Fig. 1), despite them both carrying the 2NS resistance15,16. The second isolate (NO6047 + AVR-Rmg8) (henceforth referred to as NO6047 + AVR8) is a derivative of isolate NO6047 that was transformed with allele eI of AVR-Rmg8 under control of the PWL2 promoter17 (Supplementary Table 2). NO6047 contains an alternative allele of the AVR-Rmg8 effector, designated eII′′′, which is not recognized by Rmg813. This makes NO6047 an ideal isolate to host the effector AVR-Rmg8. Isolate NO6047 is virulent on wheat line S-615, which carries Rmg8, whereas isolate NO6047 + AVR8 is not virulent because of host recognition of AVR-Rmg817. The resistance of wheat accessions showing resistance to NO6047 + AVR8 but not to NO6047 were assumed to be due to recognition of the AVR-Rmg8 effector.
We screened seedlings of a panel of 320 wheat lines including 300 landraces from the A. E. Watkins collection (Wingen et al.18) and wheat lines with chromosome-scale assemblies15 using the two MoT isolates described above (Supplementary Table 3). Only 13 accessions were highly resistant to the NO6047 + AVR8 isolate (score of 1.5 or less) (Supplementary Table 4 and Supplementary Fig. 2), only 10 accessions were highly resistant to Py 15.1.018 (score of 1.5 or less) and 9 of these were also highly resistant to both isolates suggesting that, while resistance is rare, the majority of the resistance observed was due to the recognition of the same effector in the two isolates (Fig. 1b, Supplementary Table 4 and Supplementary Fig. 3). The accessions that conferred resistance to only NO6047 + AVR8 likely contain an additional resistance that recognizes an effector absent in Py 15.1.018. Of the cultivars with chromosome or scaffold-scale assemblies, three (SY-Mattis, CDC Stanley, Claire) were highly resistant to both isolates (but susceptible to the wild-type isolate NO6047) (Supplementary Figs. 4 and 5). The resistance of these three cultivars enables them to be used as references for subsequent analyses to locate and identify the causal gene.
a, k-mers (WATDE0310) associated with resistance to Py 15.1.018 mapped to the SY-Mattis genome. Points on the y axis depict k-mers positively associated with resistance in blue. Point size is proportional to the number of k-mers. The association score is defined as the −log10 of the P value obtained using the likelihood ratio test for nested models (two-sided). The ideogram shows the position of the A. ventricosa 2NS segment (orange) and the AgRenSeq 2A association (blue) on the distal ends of the short and long chromosome arms, respectively. b, k-mer-based phylogeny of wheat landraces showing the phenotype of an accession after inoculation with Py 15.1.018. The phenotype of an accession after inoculation is indicated by the colour used to highlight the label of that accession (green, resistant (scores less than or equal to 3); yellow, intermediate (scores more than 3, less than 5) and orange, susceptible (scores equal to or greater than 5). Black circles indicate the presence of the chromosome 2A peak based on the AgRenSeq association plots. c, Representative cluster heat map for the haplotypes within the chromosome 2A interval using SY-Mattis as the reference. The phenotype of an accession after inoculation with Py 15.1.018 is indicated by the colour used to highlight the label of that accession, as in b. The darker the colour within a 50 kb window the more identical by state that sequence is to SY-Mattis. Among the accessions carrying the chromosome 2A interval, two regions were particularly similar to SY-Mattis, region 1 (788,550,000 to 789,550,000) and region 2 (793,250,000 to 794,250,000). Note that the region 1 haplotype block extends approximately 250 kb upstream of the 5.3 Mb 2A interval. Flame, Claire, Riband, Shango, WATDE0102, WATDE0171 and WATDE0310 were resistant but only contained region 1. WATDE0056 and WATDE0720 were susceptible and lacked the first 400 kb of region 1, highlighted in a yellow box. Source data are available in ref. 69. d, Gene content of the 400 kb in Mattis according to the de novo gene models. Genes coloured orange and grey correspond to high- and low-confidence genes, respectively. Genes that show expression in total leaf tissue at the three-leaf stage are labelled with their gene codes. e, Wheat blast detached leaf and spike assays for the Pm4b EMS-induced mutants of Fed-Pm4b and Pm4b over-expressors in the Bobwhite S26 background. Leaves and spikes were inoculated with M. oryzae isolates Br48ΔeI and Br48ΔeI+eI at 22 °C, denoted by ‘−’ and ‘+’, respectively.
The NO6047 + AVR8 Watkins phenotype data were analysed using NLR-enriched k-mer based association genetics (henceforth referred to as AgRenSeq14) with SY-Mattis as the reference genome. This produced a clear association peak on chromosome arm 2AL, spanning 788.8 Mbp to 794.1 Mbp. The Py 15.1.018 leaf disease scores were also analysed using SY-Mattis as the reference and produced an identical association peak (Fig. 1a). Ten Watkins accessions highly resistant to Py 15.1.018 produced associations within the same interval on chromosome arm 2AL using SY-Mattis as the reference (Fig. 1a, Supplementary Figs. 6 and 7 and Supplementary Table 4). These data mapped the resistance to AVR–Rmg8 to a 5.3 Mbp chromosome 2A interval on SY-Mattis.
Interrogating the 5.3 Mbp chromosome 2A interval
Haplotype analysis was run across the genomic sequence of the AVR–Rmg8 resistance interval in the full Watkins collection (827 accessions) and a selection of modern wheat varieties (218 cultivars) using SY-Mattis as the reference genome14,19. A cluster heat map with 50 kb window size was generated to identify regions identical or near-identical to SY-Mattis which revealed that an additional 20 accessions of the Watkins collection carry sections of the 5.3 Mbp interval. Within the 5.3 Mbp interval, two 1 Mbp blocks of similarity among 63 accessions were observed, ‘region 1’ (788,550,000 bp to 789,550,000 bp) and ‘region 2’ (793,250,000 bp to 794,250,000 bp) (Fig. 1c). These 20 Watkins accessions, along with an additional 10 lines that lack the AVR–Rmg8 resistance interval, were phenotyped with isolates NO6047 + AVR8 and Py 15.1.018 (Supplementary Table 5). A cluster heat map containing the 20 additional Watkins lines is shown in Supplementary Fig. 8. Interestingly, the wild tetraploid wheat accession 33255 was highly similar to all of region 1 and to 600 kb of region 2, indicating that the interval may have originated from a hybridization between hexaploid wheat with a wheat wild relative similar to Triticum turgidum (Fig. 1c and Supplementary Fig. 8)15. Five Watkins lines (WATDE0102, WATDE0171, WATDE0310, WATDE0566 and WATDE0804) contained only region 1 but showed resistance, suggesting that the AVR–Rmg8 resistance was contained within this interval (Fig. 1c). In addition, 22 modern European wheat cultivars also possessed only the region 1 SY-Mattis haplotype (Supplementary Fig. 8). These 22 cultivars were all resistant to both NO6047 + AVR8 and Py 15.1.018 isolates (Supplementary Table 6) confirming that the AVR–Rmg8 resistance is within the 1 Mbp interval termed region 1. Among the Watkins lines containing region 1, two lines (WATDE0056 and WATDE0720) both lacked the proximal 400 kb of region 1, and both were susceptible to NO6047 + AVR8 and Py 15.1.018 isolates indicating that the AVR–Rmg8 resistance lies within this 400 kb region (788550000–788950000 bp).
The 400 kb AVR–Rmg8 resistance interval within region 1 contains ten annotated genes (Fig. 1d and Supplementary Table 7), only five of which were expressed in RNA sequencing (RNA-seq) data from total leaf tissue at the three-leaf stage (Supplementary Table 8). Notably, four lines (WATDE0048, WATDE0527, WATDE0568 and WATDE0592) classified as containing region 1 were susceptible to both isolates. We therefore compared the sequences of the five expressed genes between the four susceptible lines and resistant lines using the whole genome Watkins sequencing data and SY-Mattis as the reference19. Two of the genes (TraesSYM2A03G00828410 and TraesSYM2A03G00828450) were monomorphic among susceptible and resistant lines, and two (TraesSYM2A03G00828400 and TraesSYM2A03G00828460) had polymorphisms which did not associate with the resistance phenotype (Supplementary Table 9). In the remaining gene (TraesSYM2A03G00828360), two of the susceptible accessions (WATDE0568 and WATDE0592) contained a T/A single-nucleotide polymorphism (SNP) converting amino acid 446 from tryptophan to a stop codon (W446*) (Supplementary Table 10), while the other two susceptible accessions (WATDE0048 and WATDE0527) possessed an identical G/A SNP converting amino acid 50 from alanine to glutamic acid (A50E). The sequences of the five non-expressed genes in the interval were also examined, and no polymorphisms that segregated with resistance were identified. Thus, the combined haplotype and allelic diversity analyses identified TraesSYM2A03G00828360 as a strong candidate gene for recognizing and conferring resistance to isolates of MoT carrying AVR-Rmg8.
The powdery mildew resistance gene Pm4 confers resistance to MoT
The RNA-seq data revealed that TraesSYM2A03G00828360 is alternatively spliced resulting in two potential transcripts. The intron/exon structure for the first five exons was the same in both transcripts, while the last exons were distinct. Transcript 1 produced a protein of 560 amino acids, while in transcript 2 the fifth intron extended an additional 1,082 bp (encapsulating the sixth exon from transcript 1) and produced a 747-amino-acid protein. BLAST analysis of transcript 1 revealed it to be almost identical to the previously reported chimeric protein of a serine/threonine kinase and multiple C2 domains and transmembrane regions that function as the wheat powdery mildew (Blumeria graminis f. sp. tritici (Bgt)) race-specific resistance gene Pm420. This study also established that Pm4 has alternate splicing, producing ‘isoforms’ Pm4b-V1 (560 amino acids) and Pm4b-V2 (747 amino acids), corresponding to TraesSYM2A03G00828360 transcript 1 and transcript 2, respectively. Both isoforms are required to confer resistance to wheat mildew20. This suggests that the wheat blast AVR–Rmg8 resistance is encoded by Pm4.
To confirm recognition of AVR-Rmg8 by Pm4, we used the germplasm resources previously developed to characterize its role in resistance to powdery mildew. This included near-isogenic lines (NILs) for two functionally distinct Pm4 alleles (Pm4a and Pm4b) in the susceptible wheat cultivar Federation (Fed-Pm4a, Fed-Pm4b), Pm4b Ethyl methanesulfonate (EMS)-induced mutants in Fed-Pm4b, and transgenic lines of susceptible cultivar Bobwhite S26 overexpressing Pm4b (Supplementary Table 11)20. To relate differences in response specifically to the presence or absence of the eI allele of AVR-Rmg8, we inoculated this germplasm with isogenic transformants of MoT isolate Br48 differing in the presence of only AVR-Rmg8. Isolate Br48∆eI has been disrupted to remove AVR-Rmg8 eI, while this gene has been replaced in isolate Br48∆eI+eI21. Federation was susceptible to both Br48∆eI and Br48∆eI+eI, while Fed-Pm4b and Fed-Pm4a (carrying different Pm4 alleles) were both resistant in seedling leaves to Br48∆eI+eI (Fig. 1e and Supplementary Fig. 9). All three lines were susceptible in spikes inoculated and incubated at 22 °C to both Br48∆eI and Br48∆eI+eI indicating that these alleles (Pm4a and Pm4b) only function in seedling resistance (Fig. 1e and Supplementary Fig. 9). Loss of wheat blast resistance in adult plants of many wheat varieties has been observed previously, but the reasons for tissue- or stage-specific resistance is unknown22.
All eight loss-of-function EMS-induced mutants of Fed-Pm4b were susceptible to both MoT isolates in seedling assays (Fig. 1e and Supplementary Figs. 9 and 10). Mutations were present in exon 6 and 7 specific to Pm4b_V1 and Pm4b_V2, respectively, indicating that both transcripts are required for resistance to MoT as was found to be the case for Bgt20. While Bobwhite S26 and lines S#3 and S#52 segregating from the T1 plants but lacking the transgene were susceptible to Br48∆eI and Br48∆eI+eI in seedling assays, both Pm4b over-expressing lines (Nr#3 and Nr#52) carrying the full-length complementary DNAs of Pm4b_V1 and Pm4b_V2 were resistant to Br48∆eI+eI. Surprisingly, Nr#52 was resistant to Br48∆eI+eI in spikes, while Nr#3 was susceptible (Fig. 1e). Sánchez-Martín et al.20 report that Nr#3 contains single copies of Pm4b_V1 and Pm4b_V2, while Nr#52 contains two or more copies of the transcripts. Expression of Pm4b_V1 and Pm4b_V2 was assessed in spike tissues of Nr#3 and Nr#52. Expression of Pm4b_V1 was significantly higher in Nr#52 compared to Nr#3 (P AVR-Rmg8 (Supplementary Fig. 11). The requirement for multiple copies or high expression of genes to provide full resistance has been reported recently23, suggesting that increased copy number, or expression levels, may provide a route to increase disease resistance.
Allelic variation
We designed PCR-based assays to detect Pm4 and used these to investigate its prevalence in landraces and modern adapted varieties (primers ‘P1_F_hex’, ‘P1_F_fam’ and ‘P1_COM’ detailed in Supplementary Table 12). These primers do not differentiate between the different Pm4 alleles. We found Pm4 to be uncommon among the landraces within the Watkins collection, being present in only 28 of 827 (3.4%) accessions. The proportion of Pm4-containing varieties was higher (15.5%; 67 out of 432) in the ‘Gediflux’ collection of highly successful European varieties from the period 1945–2000 (ref. 18) (Supplementary Table 13). This probably reflects the selection of Pm4 by breeders in Europe to control mildew while this disease is of lesser importance in many other parts of the globe.
An allelic series of Pm4, each recognizing different isolates of Bgt, has been reported, and many of these originate from wild relatives of T. aestivum20. Pm4b/Pm4c share 100% nucleotide sequence identity, as do Pm4d/Pm4e, and are henceforth referred to as Pm4b and Pm4d, respectively24. Pm4a and Pm4b were introduced from tetraploid wheats25,26, while Pm4d is believed to have been introgressed from Triticum monococcum27. The origins of Pm4f, Pm4g and Pm4h are unknown. The two alleles of Pm4 identified within this study (A50E and W446*) had sequences most similar to Pm4f but have not been reported previously. We designated these as Pm4i and Pm4j, respectively (Fig. 2b). Pm4b was the most common Pm4 allele among the Pm4-containing modern wheat varieties included in the haplotype analysis that were genotyped (71%), but it was rare among the Watkins collection (11%). By contrast, Pm4f was not observed in the modern varieties but was present in 21 of the 28 Watkins collection accessions containing Pm4 (Supplementary Table 14). Pm4d was absent within the Watkins collection and was only found in combination with the 2NS segment (33 Mbp, ref. 28) on the short arm of chromosome 2A introgressed from A. ventricosa into the wheat cultivar VPM129 (Supplementary Table 14). The 2NS segment on the short arm of chromosome 2A carries the rust resistance genes Sr38, Yr17 and Lr37 along with resistance to isolates of MoT5. It has been proposed that Pm4 was introduced into the long arm of chromosome 2A from the Triticum persicum parent of VPM130. The absence of Pm4d in Watkins accessions and its presence alongside the 2NS segment from A. ventricosa in modern varieties supports this view and indicates that the Pm4d allele may have been introduced only once into T. aestivum through VPM1 at the same time as 2NS but at the opposite end of the 2A chromosome and from a different wheat relative. This represents a second example of the serendipitous introduction of resistance into wheat from VPM1 as this line was originally developed to introduce the Pch1 eyespot resistance gene on chromosome 7Dv of A. ventricosa into wheat31, and the presence of the 2NS on the end of the short arm of chromosome 2A was not recognized.
a, Representative wheat blast detached leaf and spike assays for the Pm4b, Pm4d, Pm4f, Pm4i and Pm4j alleles, inoculated with Py 15.1.018 at 22 °C. b, Protein sequence comparison of the known Pm4 alleles. Dots represent the same amino acid present in Pm4a. c, Representative wheat blast detached leaf assays for the known Pm4 alleles, inoculated with M. oryzae isolates Br48ΔeI, Br48ΔeI+eI, Br48ΔeI+eII and Br48ΔeI+eII′ at 22 °C.
The efficacy of seedling and spike resistance against AVR-Rmg8 (isolate Py 15.1.018) was compared at 22 °C and 26 °C across wheat accessions carrying different Pm4 alleles as it has been reported that resistance to MoT is often temperature sensitive6. Carriers of Pm4b, Pm4d and Pm4f were all resistant at the seedling stage at both temperatures (Fig. 2a and Supplementary Fig. 12). These three Pm4 alleles, however, differed in efficacy in spikes. Carriers of Pm4d and Pm4f were resistant at 22 °C, while Pm4b carriers were susceptible confirming the ineffectiveness of the Pm4b allele observed in the Fed-Pm4b NIL (Figs. 1e and 2a). The level of expression of the V1 and V2 transcripts of Pm4b and Pm4f in spikes were not significantly different among the wheat varieties examined (P ≥ 0.314 and P ≥ 0.750 or V1 and V2 transcripts, respectively) indicating that differences in resistance more probably reflect differences in interaction between host and pathogen components than differences in expression of Pm4 (Supplementary Fig. 11). Carriers of Pm4d expressed moderate resistance in the spikes at 26 °C, while carriers of Pm4b and Pm4f were susceptible at this temperature. It should be noted that all the carriers of Pm4d also contained the 2NS segment that functions only in spike tissues5, and the resistance may reflect the presence of the two resistances in these varieties.
The effectiveness of Pm4 alleles against alleles of AVR-Rmg8
Three alleles of AVR-Rmg8 (eI, eII and eII′) were identified among MoT isolates collected in Brazil with eII being predominant32. The clonal lineage present in Bangladesh and Zambia, however, contains the eI allele21. Isolates transformed to carry different alleles of AVR-Rmg8 (eI, eII and eII′) differed in aggressiveness towards a wheat line (IL191) carrying Rmg8, with resistance being more pronounced against isolates carrying eI than those carrying eII or eII′ (ref. 21).
The relative effectiveness of Pm4 alleles against different alleles of AVR-Rmg8 was examined by screening seedlings of wheat lines carrying different Pm4 alleles for resistance to transformants lacking AVR-Rmg8 or carrying eI, eII or eII′ alleles (Supplementary Table 15)21. The majority of Pm4 alleles conferred resistance to all three AVR-Rmg8 effector alleles: Pm4a, Pm4b, Pm4d, Pm4f, Pm4h and Pm4i (Fig. 2c). Alleles Pm4g (accession WW-47033) and Pm4j did not confer resistance against any of the three AVR-Rmg8 effector alleles. The lack of effectiveness of Pm4j was expected as this protein is truncated and Pm4g was previously reported to be a susceptible Pm4 allele with respect to resistance to Bgt20. Interestingly, the same study also reported Pm4f to be a Bgt-susceptible allele, but it is effective against the three alleles of AVR-Rmg8. Pm4a, Pm4b, Pm4d and Pm4h (accession WW-474, ref. 33) are also highly effective against the three alleles of AVR-Rmg8. The two accessions carrying Pm4i (WATDE0048 and WATDE0527) showed greater resistance against Br48 carrying the eI or eII allele of AVR-Rmg8 than against the same isolate carrying the eII′ effector allele (Fig. 2c and Supplementary Table 15). A comparison of resistance responses of the different Pm4 alleles against selected MoT and Bgt isolates is shown in Supplementary Table 16. Differences in aggressiveness of isolates carrying different AVR-Rmg8 alleles has also been noted previously21. These two accessions, however, were highly susceptible to Py 15.1.018 (eI) and NO6047 + AVR8 (eI and eII′′) (Fig. 2a and Supplementary Fig. 13). We postulate that this may be due to the presence of additional effectors in these isolates that suppress Pm4i alleles in an equivalent manner to that reported for PWT434.
The pandemic clonal lineage of MoT present in Bangladesh and Zambia contains the eI allele of AVR-Rmg813, and so it was important to demonstrate whether Pm4 alleles would function against this lineage. Furthermore, as the impacts of MoT infection are most dramatic for spike disease, we screened spikes of wheat accessions carrying different Pm4 alleles for resistance to the Bangladesh isolate BTJP4-113. As anticipated from the studies above, accessions carrying Pm4b did not show resistance to BTJP4-1 in spikes. By contrast, spikes of accessions carrying Pm4f were highly resistant to this isolate (Fig. 3). Spikes of accessions carrying Pm4d showed moderate resistance to BTJP4-1, but this probably reflects the presence of the 2NS segment in all these accessions. Assessment of the International Maize and Wheat Improvement Center(CIMMYT)’s international screening nurseries has revealed that the 2NS segment contributes almost all the wheat blast resistance present within both the Bread Wheat Screening Nurseries and the Semi-Arid Wheat Screening Nurseries8. These authors emphasized the urgent need to identify additional non-2NS sources of resistance. We believe that the Pm4f allele provides such a source.
Wheat blast detached spike assays for Pm4b, Pm4d and Pm4f alleles inoculated with Bangladeshi isolate BTJ4P-1 at 22 °C.