An ancestral NB-LRR with duplicated 3′UTRs confers stripe rust resistance in wheat and barley

Wheat stripe rust, caused by Puccinia striiformis f. sp. tritici (Pst), is a global threat to wheat production. Aegilops tauschii, one of the wheat progenitors, carries the YrAS2388 locus for resistance to Pst on chromosome 4DS. We reveal that YrAS2388 encodes a typical nucleotide oligomerization domain-like receptor (NLR). The Pst-resistant allele YrAS2388R has duplicated 3’ untranslated regions and is characterized by alternative splicing in the nucleotide-binding domain. Mutation of the YrAS2388R allele disrupts its resistance to Pst in synthetic hexaploid wheat; transgenic plants with YrAS2388R show resistance to eleven Pst races in common wheat and one race of P. striiformis f. sp. hordei in barley. The YrAS2388R allele occurs only in Ae. tauschii and the Ae. tauschii-derived synthetic wheat; it is absent in 100% (n = 461) of common wheat lines tested. The cloning of YrAS2388R will facilitate breeding for stripe rust resistance in wheat and other Triticeae species.

W heat (Triticum spp.) is the largest acreage crop in the world. With an approximate 220 million hectares and 760 million tons in 2018, wheat was ranked second in global production after maize 1 . As a staple food crop, wheat provides about 20% of global calories for human consumption 2 . Because the world population is projected to increase by nearly two billion people within the next three decades 3 , the increasing human population worldwide will place an even greater demand for wheat production globally.
Wheat stripe rust (or yellow rust; abbreviated as Yr), caused by Puccinia striiformis f. sp. tritici (Pst), is a serious fungal disease that poses a huge threat to wheat production in regions with cool and moist weather conditions 4 , including major wheat-producing countries, such as Australia, Canada, China, France, India, the United States, and many others 5,6 . Planting wheat cultivars with adequate levels of resistance is the most practical and sustainable method to control stripe rust. Host resistance of wheat against Pst is normally classified as either all-stage resistance (ASR) or adultplant resistance (APR). Whereas ASR is effective starting at the seedling stage through the late stages of plant growth, APR is mainly effective at the late stages of plant growth 7 . In wheat, ASR confers high levels of resistance to specific Pst races, but the underlying genes, such as Yr9 8 and Yr17 9 , are often circumvented by the emergence of new virulent races. In contrast, APR typically provides a partial level of resistance, but is more durable and is effective against all or a wider spectrum of Pst races than ASR. High-temperature adult-plant (HTAP) resistance is a major type of APR; HTAP typically provides durable and non-race-specific resistance to Pst 10 . Incorporating multiple ASR and HTAP genes appears to be an excellent strategy for maintaining sustainable resistance to wheat stripe rust 10 .
Over 80 wheat stripe rust resistance (R) genes (Yr1-Yr81) have been permanently named 11 . Of the seven genes cloned so far, Yr5, Yr7 and YrSP, a gene cluster, encodes nucleotide-binding (NB) and leucine-rich repeat (LRR) proteins 12 ; Yr15 has two kinaselike domains 13 ; Yr36 has a kinase domain and a lipid binding domain 14 ; Lr34/Yr18 encodes a putative ABC transporter 15 ; and Lr67/Yr46 encodes a predicted hexose transporter 16 . While the Yr5/Yr7/YrSP cluster and Yr15 confer ASR resistance to wheat stripe rust, Yr18, Yr36, and Yr46 confer APR or HTAP resistance. Of the three cloned APR genes, only the Yr18 gene has been widely used in wheat cultivars 17 ; however, Yr18 alone does not confer adequate resistance under high disease pressures. Yr7 and YrSP confer high levels of resistance, but Pst races virulent to Yr7 are common globally and those virulent to YrSP occur in some countries 18 . Yr5 and Yr15 confer high levels of resistance to a wide range of Pst races 12,19 , but the increasing adoption of them in wheat cultivars may cause the emergence of virulent races. Characterization of additional R genes is essential in order to assemble effective resistance to constantly changing populations of Pst.
Aegilops tauschii Coss. (2n = 2 × = 14, DD) is the D genome progenitor of common wheat 20,21 . The diverse Ae. tauschii D genome offers a valuable gene pool for stripe rust resistance 20,22 . To date, several stripe rust resistance genes have been mapped in Ae. tauschii, including YrAS2388 23 and Yr28 24 on 4DS 22,25 , and YrY201 26 on 7DL. Synthetic hexaploid wheat (SHW) lines, which contain a diversity of Ae. tauschii accessions 27 , are potential breeding stocks. However, many biotic and abiotic resistance genes are suppressed in the hexaploid background 28 . To prevent a linkage drag of undesirable traits and resistance suppression, it is best to identify R genes and use them precisely in gene pyramids. In this study, we have cloned the stripe rust resistance gene YrAS2388 from Ae. tauschii. Additionally, we have demonstrated that this gene can express effectively in hexaploid wheat and barley. Deployment of YrAS2388R in wheat cultivars together with other effective genes should sustainably protect wheat production from the devastating disease wheat stripe rust.
Haplotype markers indicated that NLR 4DS-1 is YrAS2388. To help identify the correct gene, we genotyped 159 Ae. tauschii Scale bar = 1 cm. b Genetic maps are based on popC-1 (upper) and popA-2 (lower). c Physical maps are based on three fosmid clones: F2-1 (= PC1104), Fe-19 and Fa-13, which contain five genes (colored rectangles; arrows pointing to 3′ ends) that encode for two 4-helix bundle-nucleotide binding-leucine rich repeat (NLRs), a malectin-like kinase (MLK), and two receptor-like kinases (RLKs). A 3.9-kb physical gap between Fe-19 and Fa-13 was closed by sequencing PCR clones. d Genomic structure of NLR 4DS-1 in PI511383. The conserved domains and the duplicated 3′UTRs are labelled; their approximate genomic locations are highlighted with dotted lines. The 3′UTR duplication was caused by a 2668-bp insertion (magenta region), which has three regions (with a prime symbol) similar to exons 5, 6, and 7. Introns 7a and 7b, and exon 8 are the original components of the ancestral 3′UTR, but the 2668-bp insertion disrupted the ancestral 3′UTR and then formed two 3′UTRs, each containing both ancestral (black dots) and inserted (magenta dots) segments. The cryptic intron in exon 5 is highlighted by a gray box. Introns 7a, 7a′, and 7b have a size bar below their names. e Transcript variants of NLR 4DS-1 in accession PI511383. Cloning and sequencing of the NLR 4DS-1 cDNA clones identified five transcript variants, designated TV1 to TV4b, of NLR 4DS-1 in accession PI511383. Grey boxes indicate portions of the retained intron in mature messenger RNA. Rectangles and straight lines indicate regions present in mRNA; the caret-shaped lines represent regions that are absent in mRNA. Part of the cryptic intron in exon 5 is retained in TV3. TV4a and TV4b encode an identical protein, called TV4. P197 and P198 are primers that detect all five splicing variants in one PCR. Abbreviations include exon (E), four-helical bundle (4HB), intron (In), leucine-rich repeat (LRR), two miniature inverted-repeat transposable elements (M1 and M2), nucleotide-binding (NB), and start (downwards arrows in blue) and stop (downwards arrows in magenta) codons. Source data of Fig. 1a are provided as a Source Data file accessions using five markers for NLR 4DS-1 (HTM3a to HTM3e, or collectively called HTM3S), one for RLK 4DS-1 (HTM1a) and one for RLK 4DS-2 (HTM2a) (Supplementary Tables 2 and 3, Supplementary Data 3). The R-type allele (e.g. "A" in PI511383) of NLR 4DS-1 was completely associated with Pst resistance in resistant haplotypes R1 to R3 (Supplementary Data 3). All non-A scores of the NLR 4DS-1 markers were associated with Pst susceptibility. The coding region (ATG to 3′UTR2; Fig. 1d) of NLR 4DS-1 is identical amongst eight Pst-resistant Ae. tauschii accessions, including AS2386, AS2387, AS2399, AS2402, CIae9, PI349037, PI511383, and PI511384. In contrast, in RLK 4DS-1 and RLK 4DS-2 , the R-type allele (e.g. "A" in PI511383) was present in the Pst-susceptible genotypes (S1-S3 and S5), indicating that both genes do not confer Pst resistance. Similarly, the absence of RLK 4DS-1 and/or RLK 4DS-2 in the R2 and R3 haplotypes suggested that neither gene is essential for Pst resistance. Thus, NLR 4DS-1 is the only candidate for YrAS2388R.   Fig. 2c). T 1 plants with functional NLR 4DS-1 , RLK 4DS-1 and RLK 4DS-2 were resistant (IT scores = 1-5), while the ones lacking the three genes were susceptible (IT scores = 7-8). Therefore, the fosmid PC1104 confers stripe rust resistance in transgenic wheat and barley.
The Pst-resistant NLR 4DS-1 has duplicated 3′UTRs ( Fig. 1d) in all Pst-resistant parents (CIae9, PI511383 and PI511384) and each 3′UTR is associated with multiple transcript variants: TV1 and TV2 with 3′UTR1, TV3 and TV4a (and 4b) with 3′UTR2 ( Fig.  1e). We overexpressed the Pst-resistant NLR 4DS-1 cDNA under the maize Ubi promoter (Supplementary Table 4). All 36 transgenic wheat and barley lines that expressed TV1 (or TV2) did not confer resistance to stripe rust ( Table 2), suggesting that one cDNA isoform was insufficient to confer stripe rust resistance. For stripe rust resistance, the NLR 4DS-1 gene may require the activity of multiple cDNA isoforms and/or regulatory elements in the genomic sequence.
Innate and external factors regulate NLR 4DS-1 expression. In the Pst-resistant NLR 4DS-1 , the most abundant isoforms are TV1 (for a 1068-aa protein with complete 4HB, NB, and LRR domains) and TV4 (for a 471-aa protein with a complete 4HB and a partial NB domain) (Fig. 1e, Supplementary Fig. 7a). The less abundant isoform TV2 might result from either a partial exon skipping from TV1 or the retention of an 833-bp cryptic intron in exon 5, which disrupts the NB and LRR domains. TV3 is also a less abundant isoform and is structurally similar to TV4, but retains the first 244 bp in the 833-bp cryptic intron, which only disrupts the LRR domain. In contrast, the Pst-susceptible NLR 4DS-1 either remained completely silent in PI486274 and PI560536 or produced only the TV1-type transcript in AL8/78 and AS87 (Supplementary Fig. 3b).
In Pst-resistant PI511383, TV1 to TV4 cDNAs were all expressed in the seedling and adult leaves ( Supplementary Fig. 3c). When exposed to alternating low (10°C) and high (25°C) temperatures, the high temperature upregulated TV2 and downregulated TV4 ( Supplementary Fig. 3d), which is correlated with increased Pstresistance at elevated temperatures. In response to Pst race PSTv-306, the TV1 cDNA levels in the Pst-infected plants were comparable to those in the mock-inoculated control plants ( Supplementary Fig. 3e). In contrast, Pst infections upregulated TV2 at 2, 5, and 10 days post inoculation (dpi) but not at 3 dpi, downregulated TV3 at 3, 7, and 14 dpi, and downregulated TV4 at 3 dpi ( Supplementary Fig. 3e). Thus, both temperature and Pst infection regulate the transcription of NLR 4DS-1 . However, a change in the relative levels of either the individual four transcripts and/or the proteins or protein complexes may affect the induction of stripe rust resistance. Among the Pst-susceptible mutations of NLR 4DS-1 , Ser394Asn and Gln557Stop(*) only affect TV1 and Thr456Ile only affects TV4, which indicates that both  TV1 and TV4 are essential for stripe rust resistance (Supplementary Data 4). Collectively, we hypothesize that TV1 plays a major role in the induction of stripe rust resistance, TV2 acts as a positive co-factor, and TV4 (or possibly TV3) act either as negative regulators when its expression is high or as positive regulators when its expression is low (Supplementary Fig. 8).
Using a yeast two-hybrid system, we tested the interaction among the native (TV1, TV2, and TV4) and mutant (TV1 G117D , TV2 G117D and TV2 V267I ; Supplementary Data 4, Supplementary  Fig. 7a) isoforms of the Pst-resistant NLR 4DS-1 . The NLR 4DS-1 isoforms, both native and mutant forms (NM forms) had no autoactivity. A strong interaction occurred amongst the TV2 proteins (NM forms; Supplementary Fig. 7b). We observed a weak interaction between TV2 mutants and TV1 (NM forms), and between TV2 proteins (NM forms) and TV4. Apparently, TV2 can mediate protein interactions amongst multiple isoforms of NLR 4DS-1 . The Pst-resistant NLR 4DS-1 may arise from paralogous genes. All Pst-resistant NLR 4DS-1 genes contain two duplicated regions. The first region includes the 3′ end of exon 5, exons 6 and 7, and intron 7a; and the second region includes the pseudo-exon 5′, exons 6′ and 7′, and intron 7a′ (Fig. 1d, e, Supplementary Fig. 9a). This duplication is not present either in Pst-susceptible NLR 4DS-1 alleles or in any NLR 4DS-1 -like genes. To examine the origin of the duplicated regions, we built separate phylogenetic trees for each of six selected fragments (exons 5, 6, 7, and 8; and introns 7a and 7b) of 7 to 15 NLR 4DS-1 homologues in Triticeae (Supplementary Fig. 9b). The trees indicate that exons (5-8) and introns (7a and 7b) of the Pst-resistant NLR 4DS-1 are more related to those of the Pst-susceptible NLR 4DS-1 in CS (CS-4D:1821950..1825589); all the duplicated fragments (exons 5′ to 7′ and intron 7a′) are in separate clades. In addition, the duplicated 3′UTR1 and 3′UTR2 DNA of NLR 4DS-1 in PI511383 are only 87% identical in the conserved 373 bp (GenBank MK736661: 3735..4107 versus 6409..6781, counted forward from the start codon ATG). Thus, the Pst-resistant NLR 4DS-1 likely arose after a shuffling event between two paralogous genes. Specifically, the 3′UTR2 contains part of a 2668-bp insertion (within a 6-bp target site duplication = TACTGG) that occurred in intron 7 of the ancestral 3′UTR1 region. A similar 3′UTR duplication in the Pst-resistant NLR 4DS-1 gene is present in the synthetic wheat W7984 35 . In the 2668-bp insertion, a 496-bp region (pseudo-exon exon 5′) is 90% identical to the ancestral exon 5. The insertion also has two miniature inverted-repeat transposable elements, which are frequently adjacent to transcriptionally active genes 36 . Likely, the 2668-bp fragment was derived from another, currently unidentified, NLR 4DS-1 homologue in Ae. tauschii.
In Triticeae, there are multiple NLR 4DS-1 -like genes; three copies were identified in the YrAS2388 region ( Supplementary Fig. 2). In common wheat CS, there are at least five transcriptionally active homologues of the NLR 4DS-1 gene (Supplementary Fig. 10). None of the NLR 4DS-1 -like homologues in CS has duplicated 3′ UTRs. The Pst-susceptible NLR 4DS-1 homologues in CS share only 86%-94% identity with the Pst-resistant NLR 4DS-1 in PI511383 at the cDNA level.
NLR 4DS-1 offers a toolbox for solving stripe rust problems. We compared the stripe rust resistance in 81 SHW lines 33 and their original parents, including 30 SHW lines with the YrAS2388R gene (Supplementary Data 5). YrAS2388R confers a strong Pst resistance (IT scores = 1-3) in Ae. tauschii 22 . However, 27% of SHW wheat had significantly less resistance than the parental lines (T. turgidum and/or Ae. tauschii). In this study, SW3 has the Pst-resistant NLR 4DS-1 allele and shows the characteristic expression of alternatively spliced transcripts. However, SW3 was susceptible (IT scores = 7-9) to Pst in Moscow, ID, USA (Table 1), presumably because of a suppressor in its genetic background. Nonetheless, Ae. tauschii accessions with a strong Pst resistance frequently conferred moderate to high Pst resistance in a derived SHW wheat (Supplementary Data 5), indicating that Ae. tauschii is valuable for breeding resistant NLR 4DS-1 . For example, the SHW wheat Syn-SAU-S9 is based on Langdon/AS313//AS2399, in which the Ae. tauschii AS2399 is positive for the YrAS2388R gene 22 . Although Syn-SAU-S9 displayed only moderate resistance to Pst (IT scores = 4-5), we used Syn-SAU-S9 to transfer the YrAS2388R gene into common wheat. Three co-segregating markers were used for marker-assisted selection of YrAS2388R ( Supplementary Fig.  11, Supplementary Table 2). In 2015, we developed an elite line Shumai 1675, which is an F 6 line of Syn-SAU-S9/Chuan 07001//Shumai 969. Shumai 1675 is highly resistant to Pst in Sichuan, China. In 2017, Shumai 1675 outcompeted the check variety Mianmai 367 with an 11% increase in yield in the regional variety trials of the Sichuan province, China (Supplementary Table 6).

Discussion
YrAS2388R provides robust resistance in a wide spatial and temporal range, including China (current study), Canada 37 , Norway 25 , the United Kingdom 38 and the United States (TA2450 = CIae9, TA2452 = PI511384 39 ; current study). However, YrAS2388R has had limited use probably for two reasons: it is absent in common wheat; and it can be suppressed in hexaploid wheat. In the present study, YrAS2388R, when separated from potential linkage drag, conferred strong stripe rust resistance in transgenic wheat and barley, indicating that YrAS2388R offers a practical solution for stripe rust resistance in Triticeae. The YrAS2388R gene-based markers (e.g. Xsdauw95, Supplementary  Fig. 11) can be used for marker-assisted selection.
YrAS2388R is another example of a gene that was either not transferred or lost during domestication. Nevertheless, genes from both progenitors and distantly-related species of wheat can be used to enhance contemporary common wheat. Of the 81 permanently named Yr resistance genes, 21 were transferred from either related species or wild relatives of wheat, such as Yr5 from Triticum spelta, Yr15 and Yr36 from T. dicoccoides and Yr28 from Ae. tauschii 40 . However, alien genes can be accompanied by linkage drag 41 . For example, linked genes to Yr8 from Ae. comosa are associated with tall height and delayed maturity 42 . The Yr9 gene from the 1BL/1RS translocation improves grain yield but causes inferior quality 43 , which limits its use in wheat especially in the U.S. Pacific Northwest 44 . YrAS2388R could be transferred into wheat through a cisgenic approach. Thus, cisgenic YrAS2388R can provide an advantage to consumers in comparison to traditional breeding.
YrAS2388R (or Pst-resistant NLR 4DS-1 ) is associated with duplicated 3′UTRs, which is an apparently rare phenomenon. The ancestral 3′UTR of NLR 4DS-1 adjoined the 3′-end of an unknown NLR 4DS-1 paralog, resulting in duplicated 3′UTRs in Pst-resistant NLR 4DS-1 . The 3′UTR is an important component of eukaryotic genes 45 . More than half of human genes use alternative polyadenylation to generate mRNAs that differ in the 3′ UTR length but encode the same protein 46 . In contrast, there are few reports of genes with two separate 3′UTRs that cause a difference in the protein product. In wheat, the stripe rust resistance gene WKS1 generated six transcript variants, of which WKS1.1 differs from the others in the 3′UTRs 14 . Pst-resistant NLR 4DS-1 also shows alternative splicing (AS) in the NB-LRR region of the gene. AS is prevalent in eukaryotes 47 ; 95% of multi-exon genes in human 48 and 44% of multi-exon genes in Arabidopsis 49 display AS. In Arabidopsis, the bacterial-resistance gene RPS4 produces alternative transcripts in response to infection by pathogen Pseudomonas syringae pv. tomato 50 . Both environmental and developmental stimuli precisely regulate the abundance of functional mRNA isoforms 51 . Here, in keeping with resistance, expression of the NLR 4DS-1 isoforms also depends on pathogen infection and the temperature. Thus, abundance of NLR 4DS-1 isoforms appears to be a mechanism that wheat can use to robustly resist stripe rust pathogen invasion.
The NLR 4DS-1 protein is a member of the CC-NB-LRR (CNL) proteins. The coiled-coil domain of the potato virus X resistance protein (Rx) actually forms a four-helix bundle (4HB) 52 . The N-terminal domain of NLR 4DS-1 is predicted to fold into four helixes, and it is also classified as an Rx-CC-like in the NCBI CDD (E = 9 × 10 −9 ) and Rx_N in the Pfam database (E = 6 × 10 −16 ). Although CNL genes are often race-specific and not durable 53 , some CNL genes such as the rice blast resistance gene Pigm R 54 have been durable. Here, we showed that YrAS2388R confers resistance to a broad array of Pst races and has been effective to all natural infections of Pst in China since 1995. As a typical NLR gene, we hypothesize that the NLR 4DS-1 proteins change their state via a competition model ( Supplementary Fig. 8). The full-length TV1 protein plays a central role in signal transduction, but it requires other variant proteins (TV2 and TV4) for a proper conformation, which together form an active TV1 complex for defense signaling.
Here, YrAS2388R was fully expressed without suppression in transgenic hexaploid wheat and in barley. In addition, we have produced Shumai 1675, which has YrAS2388R and is strongly resistant to Pst, suggesting that either YrAS2388R is not suppressed in Shumai 1675 or that YrAS2388R worked positively with other Yr genes to confer resistance to Pst. However, in the current study, the resistance levels of parental lines (T. turgidum and/or Ae. tauschii) were suppressed in nearly 27% of the SHW wheat lines. Yr28, which is probably the same gene as YrAS2388 22 , was effective in seedlings and adult plants of SHW Altar 84/Ae. tauschii accession W-219 24 . Here, we observed that YrAS2388R in SHW SW3 was suppressed, i.e., it was fully susceptible to natural Pst races at adult-plant stages in Moscow (ID, USA), probably because the suppressor responds more to the cooler night temperatures in this area. When YrAS2388R is suppressed in a specific hexaploid wheat such as SW3, Pstresistance levels might be increased by disrupting the unknown suppressor, as was previously done by inactivating a suppressor of stem rust resistance 55 .
In the case of wheat powdery mildew, pyramiding of closely related NLR genes can cause dominant-negative interactions and that lead to R gene suppression 56 . For example, the Pm8 resistance gene from rye was suppressed in wheat by a susceptible allele of the wheat ortholog Pm3 57 . In the present study, the Pstresistant NLR 4DS-1 in PI511383 shares 86-94% identity with cDNA from the transcriptionally active homologues in common wheat (Supplementary Fig. 10). Thus, YrAS2388R suppression might conceivably be caused by close homologues of NLR 4DS-1 that are present in Triticeae. To test this hypothesis, in the future, one could mutagenize a SW3 line, screen for truncation mutations in the NLR 4DS-1 homologues, and test whether the homologues' mutations have any effect on stripe rust resistance. Regardless, because the transgene NLR 4DS-1 induces effective Pst resistance in hexaploid wheat, we predict that sustainable Pst resistance can be achieved with either a cisgenic strategy with Pst-resistant NLR 4DS-1 or a conventional strategy that combines both the incorporation of a Pst-resistant NLR 4DS-1 and either avoidance or inactivation of the apparently linked latent suppressor(s) from Ae. tauschii.
We developed three F 2 populations (popA: PI511383/PI486274; popB: CIae9/ PI560536; and popC: PI511384/AS87 23 ). These populations were used for preliminary and fine mapping, and popC was also used to confirm the single Mendelian inheritance of YrAS2388. In popA, we selected 11 F 2 plants that were heterozygous in the YrAS2388 region (Xsdauw2-Xsdauw36), and allowed them to self-pollinate to produce F 3 seeds. After screening 4,205 F 3 plants, we identified 467 plants with crossovers in the Xsdauw2-Xsdauw36 interval, and used them to generate a high-density map.
Stripe rust inoculum and infection assays. Wheat stripe rust tests were conducted in four institutions: Shandong Agricultural University (SDAU), Tai'an, China; Sichuan Agricultural University (SCAU), Chengdu, China; Washington State University (WSU), Pullman, USA; and University of Idaho (UI), Moscow, USA. Avocet Susceptible (AvS), Huixianhong, Mingxian 169, and/or SY95-71 were used as susceptible checks and also planted surrounding the plots to increase and spread urediniospores for adequate and uniform rust levels for reliable screening.
For winter-growth genotypes tested in greenhouses or growth chambers, seeds were vernalized in wet germination paper (Anchor Paper Co., Saint Paul, MN, USA) at 4°C in darkness for 45 d; vernalized shoots were transplanted into soil in the greenhouse and maintained at 25°C during the day and 15°C at night with 16 h photoperiod.
Infection types (IT) were recorded using a 0-9 scale 59 and the following categories: resistant (R, IT scores = 0-3), moderate reactions (M, IT scores = 4-6) that include moderate resistance (MR, IT scores = 4-5) and moderate susceptibility (MS, IT score = 6), and susceptible (S, IT scores = 7-9). IT scores were recorded 15-18 days post inoculation (dpi) when the uredinial pustules were clearly visible on susceptible plants. Responses of SHW and their parental lines to Pst are shown in Supplementary Data 5.
At SDAU, urediniospores were obtained from the Institute of Plant Protection, Chinese Academy of Agricultural Sciences, Beijing, China. Due to changes in race frequency and spore availability, different Pst races were used in different years (mixed spores of Chinese Pst races CYR29, CYR31, CYR32, CYR33, Su11 and/or Su14 during 2010 to 2012; CYR29 and CYR32 in 2013; and CYR29, CYR31, CYR32 and CYR33 in 2014-2016). Collectively, these races represent the predominant Pst races in China in different periods since the 1990's. Field trials were performed to assess the responses to Pst in the parental lines, F 1 , F 2 and advanced progenies of popA and popB. At the seedling stage, an aqueous spore suspension was manually injected with a 2.5 ml syringe into leaf bundles and repeated after 10 days. At WSU, urediniospores were produced by the USDA-ARS Wheat Unit at Pullman, WA, USA. The plants were initially grown in a greenhouse at 15 to 25°C. At the two-leaf stage, we prepared a mixture of urediniospore and talc at 1:20 ratio (v vs. v), dusted it on plants, and then applied a water mist onto the plants. The inoculated plants were incubated in a dew chamber at 10°C in the dark for 24 h, and then moved to growth chambers for either low or high temperature tests. The low temperature (LT) cycle had a 16-h photoperiod (6 a.m.-10 p.m.) with a diurnal temperature cycle of 4°C at 2 am and a gradual increase to 20°C at 2 pm followed by a gradual decrease to 4°C at 2 am. The high temperature (HT) cycle had a 16-h photoperiod with a gradual temperature gradient from 10°C at 2 a.m. to 30°C at 2 p.m. and then back to 10°C at 2 a.m..
At UI, urediniospores were produced by the USDA-ARS Wheat Unit at Pullman, WA, USA. Transgenic plants and wild-type controls were grown in chambers. At either the two-leaf stage for seedlings or at the 6-leaf stage for adults, plants were dust-inoculated using the urediniospore and talc mixture (1:20) Bulked segregant analysis of the YrAS2388 gene. Genomic DNA was extracted using the Sarkosyl method 17 . Infinium iSelect genotyping was assayed at the Genome Center (University of California, Davis, CA, USA). Normalized Cy3 and Cy5 fluorescence for each DNA sample was plotted with the GenomeStudio program (Illumina, Inc., San Diego, CA, USA), resulting in genotype clustering for each SNP marker 20 .
We performed bulked segregant analysis (BSA) on four parents (CIae9, PI486274, PI511383 and PI560536) and 17 Pst-susceptible F 2 plants, ten from popA and seven from popB (Supplementary Table 1), using the wheat 10k iSelect array 29 . The Pst responses of the tested plants were obtained in the field in 2011. For SNP data, we sequentially eliminated: (1) those with missing data or that were being heterozygous in the parents, (2) those that were being polymorphic between the two resistant parents or between the two susceptible lines, (3) those that were identical among the four parents, and (4) those with four or more missing data points amongst 17 Pst-susceptible F 2 plants. We retained 3276 SNP loci for BSA analysis. Among Pst-susceptible F 2 plants with a clear genotype, the frequency of a homozygous "B" genotype (= susceptible phenotype) was calculated and sorted in descending order for each SNP. The top 20 SNPs were prioritized for further analysis.
Preliminary and fine mapping of the YrAS2388 gene. We targeted the AT4D3406 region ( Supplementary Fig. 1a) to develop PCR markers, which was facilitated by using the Ae. tauschii SNP map 29 , and the genome sequences of Ae. tauschii 30 , common wheat (IWGSC RefSeq v1.0 60 ), synthetic wheat 35 and 20 fosmid clones of PI511383. Markers were primarily based on insertion-deletion polymorphisms (InDel), cleaved amplified polymorphic sequences (CAPS) and derived cleaved amplified polymorphic sequences (dCAPS 61 ). PCR primers, restriction enzymes and annealing temperatures are described in the Supplementary Table 2. All other oligos used in the current study are documented in the Supplementary Table 8. PCR products were separated in either 6% non-denaturing acrylamide or 2% agarose gels. The 4DS maps ( Supplementary Fig. 1) were calculated using the maximum likelihood algorithm and the Kosambi function in JoinMap 4.0 (Kyazma B.V., Wageningen, Netherlands) and were assembled using MapChart v2.3 (www.wur.nl/en/show/Mapchart.htm).
Construction and screening of the fosmid genomic library. PI511383 leaf tissue was harvested from 4-week-old plants and stored at −80°C. Megabase-size DNA was prepared by embedding nuclei in 0.5% low-melting agarose, followed by nuclear lysis in the presence of detergent and proteinase-K 62 . Sixty DNA plugs were transferred to individual 1.5-ml tubes with 200-μl TE buffer. DNA in agarose was sheared by 22 freeze-thaw cycles with incubation in liquid nitrogen for 20 s and then a 45°C water bath for 3 min. The sheared DNA in a 33.5-63.5-kb range was purified from a gel, repaired using the DNA End-Repair enzyme, ligated into the pCC1FOS vector, and packed into the phage particles as instructed by the Copy-Control™ Fosmid Library Production Kit (Epicentre Technologies Corp., Madison, WI, USA). Packaged fosmid clones were transformed into the EPI300-T1R competent cells, and the titer of the genomic library was calculated as indicated in the manual (Epicentre). On average, 1,000 or 2,000 clones per plate were obtained from a diluted solution after an 18 h to 24 h incubation at 37°C. Colonies were recovered using a mix with 6 ml LB and 1.8 ml glycerol, divided into three aliquots (2 ml each, super colony pools), and stored at −80°C.
PCR screening was performed on each of 622 super colony pools, with 2-μl bacterium stock as template. We screened for markers Xsdau93, Xsdau95 and O13 (PCR primers P160/P161) ( Supplementary Tables 2 and 8). PCR amplification was performed as follows: 95°C for 5 min, 32 cycles with 95°C for 30 s, 58°C for 30 s and 72°C for 50 s, and a final extension at 72°C for 10 min. PCR products were separated on a 1% agarose gel and visualized by ethidium bromide staining. For positive super colony pools, 25-μl glycerol stock was inoculated into 5-ml liquid LB supplemented with 12.5 μg chloramphenicol ml −1 (LB-C), and cultured on a 250 rpm shaker at 37°C for 4 h. The culture was diluted in a 10-fold series (10 −1 to 10 −5 ) using liquid LB, and the serial dilutions (300 μl per level) were plated onto the LB-C agar. An ideal dilution yielded 4,000-5,000 clones per 15-cm-diameter plate, from which colonies were collected using the 384-pin replicator with four repeated contacts to collect more representative colonies. After the replicator was used to inoculate a 384-well plate with 50-μl liquid LB-C, the plate was incubated at 37°C overnight. Each well was screened by PCR. For positive wells, 20-μl culture was enriched in 2-ml liquid LB-C, and grown in a 250 rpm shaker at 37°C for 2 h. The end culture was diluted 10-fold (from 10 −1 to 10 −4 ) using liquid LB-C, and 100-μl culture was plated onto LB-C agar. An ideal dilution yielded 50-200 clones per 9-cm-diameter plate, from which a positive clone would be revealed among 24 clones. The Pst-susceptible mutants were screened for structural variations in the coding sequence of RLK 4DS-1 , RLK 4DS-2 , and NLR 4DS-1 . Plant DNA was prepared from the flag leaf using the Sarkosyl method 17 . Mutations of the candidate genes were identified using PCR-based DNA sequencing. For RLK 4DS-1 , we divided the 2570-bp fragment into two parts: (1) exons 1-3 between P162 and P163, and (2) exon 4 between P164 and P165. For RLK 4DS-2 , we examined a 1430-bp target region of the exon 3 between P167 and P168. For NLR 4DS-1 , we divided the 6072-bp fragment into five parts: (1) promoter and exons 1-3 between P169 and P170, (2) exon 4 between P171 and P172, (3) exon 5 between P173 and P174, (4) exon 5-6 between P175 and P176, and (5) the insertion region with 3′UTR2 between P177 and P178. PCR primers are described in the Supplementary  Table 4). To overexpress the candidate genes, we cloned the cDNA copies of NLR 4DS-1 TV1 cDNA with PCR primers P181/P182 and NLR 4DS-1 TV2 cDNA with PCR primers P181/P183. We then assembled two plant expression constructs: PC1101 (Ubi:: NLR 4DS-1 TV1-cDNA ) and PC1102 (Ubi::NLR 4DS-1 TV2-cDNA ) (Supplementary Table 4). The fosmid PC1104 has no plant selection marker, and thus required cotransformation with PC174, which has the bialaphos (BAR) and hygromycin (HYG) selection markers both under the CaMV 35S promoter (Supplementary Table 4). The other two plant expression constructs (PC1101 and PC1102) have both BAR and HYG selection markers on their T-DNA, and were used for direct transformation.
Standard methods for biolistic bombardment and tissue culture of wheat were used 63 . Using an intact fosmid PC1104, we bombarded 1,590 immature embryos of CB037 and generated 24 putative transgenic plants. Using the cleaved fosmid PC1104 (Supplementary Fig. 6), we bombarded 5,790 immature embryos of CB037 and 2,013 immature embryos of Bobwhite, and generated 197 and 20 putative transgenic plants, respectively. Using the bombardment protocol for wheat 63 , we also transferred the intact fosmid PC1104 into barley Golden Promise, however the tissue culture and regeneration procedures were specific for barley 64 . We bombarded 2,200 immature embryos of Golden Promise and generated 300 putative transgenic plants.
Haplotype analysis. Haplotype analysis was performed to understand the association of haplotypes and responses to Pst and the evolution of the YrAS2388 region. Haplotype markers (HTM) were specifically designed for RLK 4DS-1 , RLK 4DS-2 and NLR 4DS-1 . Their physical locations (in Supplementary Tables 3 and  5, Supplementary Data 3) are counted from "A" in the start codon (ATG) in the genomic allele (GenBank accession number MK288012); for each marker, two periods separate the starting and ending nucleotides, and a minus sign indicates a backward count from "A" and a plus sign indicates a forward count from "A". First, 159 Ae. tauschii accessions were genotyped in Sichuan, China using seven markers: HTM1a (= RLK 4DS-1 ), HTM2a (= RLK 4DS-2 ) and HTM3a to HTM3e (= NLR 4DS-1 ) (Supplementary Table 3

, Supplementary Data 3). Second, 874
Triticeae lines were genotyped in Shandong, China using four markers: HTM1b (= RLK 4DS-1 ), HTM2b (= RLK 4DS-2 ) and HTM3f to HTM3g (= NLR 4DS-1 ). PCR primers are described in Supplementary Table 2. Markers used to genotype the Triticeae collection in Shandong were different from those used for genotyping the Ae. tauschii collection in Sichuan. Genotypes per gene per accession were not necessarily identical between the two tested collections. Thus, grouping of haplotypes should be considered separately for these two collections.
Sequence analysis. Fosmid clones were extracted using QIAGEN Large Construct Kits (QIAGEN, Germantown, MD, USA). Library preparation, highthroughput sequencing and quality control were performed by the Berry Genomics Company (Beijing, China). In brief, DNA was fragmented, endrepaired, ligated to Illumina adaptors, and separated on a 2% agarose gel to select fragments about 400-500 bp 66 . Adaptor specific primers were used to amplify the ligation products. The final library was evaluated by qRT-PCR. PE reads (150 bp) were obtained using the Illumina HiSeq2500. Sequence reads of the vector pCC1FOS and the bacterial genome were masked by the crossmatch tool in the Phrap package 67 . A de novo assembly of each fosmid was done using either SPAdes 3.12 [http://cab.spbu.ru/software/spades/] or ABySS 2.0.2 [www. bcgsc.ca/platform/bioinfo/software/abyss/releases/2.0.2]. Orientation and order of small contigs was inferred using the reference sequences of W7984 35