About Us | Help Videos | Contact Us | Subscriptions

The Plant Genome - Original Research

Genome-Wide Association Mapping Reveals Novel QTL for Seedling Leaf Rust Resistance in a Worldwide Collection of Winter Wheat


This article in TPG

  1. Vol. 9 No. 3
    unlockOPEN ACCESS
    Received: June 02, 2016
    Accepted: July 24, 2016
    Published: September 29, 2016

    * Corresponding author(s): xiangyang.xu@ars.usda.gov
Request Permissions

  1. Genqiao Liac,
  2. Xiangyang Xu *a,
  3. Guihua Baib,
  4. Brett F. Carverc,
  5. Robert Hungerd,
  6. J. Michael Bonmane,
  7. James Kolmerf and
  8. Hongxu Dongg
  1. a Wheat, Peanut and Other Field Crops Research Unit, USDA–ARS, Stillwater, OK 74075
    c Plant and Soil Science Dep., Oklahoma State Univ., Stillwater, OK 74078
    b Hard Winter Wheat Genetics Research Unit, USDA–ARS, Manhattan, KS 66506
    d Entomology and Pathology Dep., Oklahoma State Univ., Stillwater, OK 74078
    e Small Grains and Potato Germplasm Research Unit, USDA–ARS, Aberdeen, ID 83210
    f Cereal Disease Laboratory, USDA–ARS, St. Paul, MN 55106
    g Dep. of Crop Sciences & Energy Biosciences Institute, Univ. of Illinois, Urbana, IL 61801
Core Ideas:
  • The association panel used in this study may represent the largest in wheat GWAS
  • Three pathotypes, including a bulked races collected in the field, were used
  • A total of 14 QTL for leaf rust resistance were identified at a FDR of 0.01.
  • Six QTL identified in this study are novel loci in bread wheat
  • Thirty-one additional QTL were significant at a FDR of 0.05.


Leaf rust of wheat (Triticum aestivum L.) is a major disease that causes significant yield losses worldwide. The short-lived nature of leaf rust resistance (Lr) genes necessitates a continuous search for novel sources of resistance. We performed a genome-wide association study (GWAS) on a panel of 1596 wheat accessions. The panel was evaluated for leaf rust reaction by testing with a bulk of Puccinia triticina Eriks. (Pt) isolates collected from multiple fields of Oklahoma in 2013 and two predominant races in the fields of Oklahoma in 2015. The panel was genotyped with a set of 5011 single-nucleotide polymorphism (SNP) markers. A total of 14 quantitative trait loci (QTL) for leaf rust resistance were identified at a false discovery rate (FDR) of 0.01 using the mixed linear model (MLM). Of these, eight QTL reside in the vicinity of known genes or QTL, and more studies are needed to determine their relationship with known loci. QLr.stars-7AL1 is a new QTL to bread wheat but is close to a locus previously identified in durum wheat [Triticum turgidum L. subsp. durum (Desf.) Husn.]. The other five QTL, including QLr.stars-1BL3, QLr.stars-1DC1, QLr.stars-2BL1, QLr.stars-5BL1, and QLr.stars-7AS1, are likely novel loci for leaf rust resistance. The uneven distribution of the 14 QTL in the six subpopulations of the panel suggests that wheat breeders can enhance leaf rust resistance by selectively introgressing some of these QTL into their breeding materials. In addition, another 31 QTL were significantly associated with leaf rust resistance at a FDR of 0.05.


    FDR, false discovery rate; GWAS, genome-wide association study; HR, highly resistant; HS, highly susceptible; iCore, informative core; IT, infection type; LD, linkage disequilibrium; LnP(D), log probability of data; MLM, mixed linear model; MR, moderately resistant; MS, moderately susceptible; NSGC, National Small Grains Collection; PCR, polymerase chain reaction; Pt, Puccinia triticina Eriks.; QTL, quantitative trait loci; SNP, single-nucleotide polymorphism

Leaf rust, caused by Puccinia, is a major foliar disease that has afflicted wheat for thousands of years (Bolton et al., 2008). Leaf rust occurs nearly everywhere wheat is grown and typically reduces yield by 5 to 15% or more depending on crop developmental stage when the initial rust infection occurs (Samborski, 1995). Yield loss is mainly through reduction of kernel weight and kernel number per spike (Hershman, 1985).

Genetic resistance is vital to wheat leaf rust control. Much effort has been directed to identifying Lr genes. To date, 71 Lr genes have been officially designated (Lr1Lr74; Lr21 = Lr40; Lr39 = Lr41; Lr43 deleted) and 20 more temporarily named. These Lr genes are distributed on almost every chromosome. Of these genes, 52 originated from bread wheat and the others were derived from alien species, including Aegilops speltoides Tausch, Ae. tauschii Coss., rye (Secale cereale L.), T. timopheevii (Zhuk.) Zhuk., Agropyron elongatum (Host) P. Beauv., einkorn (T. monococcum L.), wild emmer wheat [Triticum dicoccoides (Korn. ex Asch. & Graebn.) Schweinf.], spelt [T. aestivum L. subsp. spelta (L.) Thell.], Ae. ventricosa Tausch, Ae. geniculate Roth, Ae. kotschyi Boiss., Ae. neglecta Req. ex Bertol., Ae. peregrine (Hack.) Maire & Weiller, Ae. sharonensis Eig, Ae. umbellulata Zhuk., slender wheatgrass [Elymus trachycaulus (Link) Gould ex Shinners], T. armeniacum (Stolet.) Nevski, and Thinopyrum intermedium (Host) Barkworth & D. R. Dewey (McIntosh et al., 2013, 2014; http://wheat.pw.usda.gov/GG2/Triticum/wgc/2013/2013-2014_Supplement.pdf). The majority of these genes confer race-specific resistance in a gene-for-gene manner, and only a few of them, such as Lr34, Lr46, and Lr68, confer race-nonspecific resistance. Race-specific Lr genes provide high levels of resistance to specific races and have been widely used in wheat breeding. However, 40 to 60 Pt races are detected each year in North America, and growing large acreages of susceptible varieties provides a reservoir for new races in the southern Great Plains of the United States (Kolmer and Hughes, 2015). The large genetic variation in pathogen populations typically makes race-specific Lr genes ineffective in a few years after their deployment in the region (Kolmer, 2005), which forces wheat breeders to find new resistance sources.

Knowing what Lr genes are present in germplasm allows for full use and effective deployment of these genes in wheat breeding. Most known Lr genes have been mapped using the biparental population mapping approach. More recently, advances in genomics have made it feasible to genotype a large collection of germplasm lines with thousands of SNP markers. As a result, the GWAS approach, based on the principle of linkage disequilibrium (LD), has been effectively used to exploit existing allelic diversity for traits of agronomic importance. Compared with biparental population mapping, GWAS may provide a higher mapping resolution because all historical meiotic events that have occurred in ancestors of a diverse panel of germplasm can be used. Moreover, GWAS enables detection of genes or QTL present in multiple germplasm lines rather than one or two parental lines of a biparental population. Thus, GWAS is an effective alternative to traditional linkage mapping. In crops with reference genome sequences available, such as rice (Oryza sativa L.) (Huang et al., 2010; 2012a,b; Chen et al., 2014), maize (Zea mays L.) (Gore et al., 2009; Buckler et al., 2009; Tian et al., 2011; Peiffer et al., 2013, 2014; Poland et al., 2011; Chia et al., 2012), and foxtail millet [Setaria italica (L.) P. Beauv. subsp. italica] (Jia et al., 2013), high-density HapMaps have been constructed. These HapMaps greatly facilitate GWAS and have the potential to rapidly resolve complex traits to gene-level resolution (Xu and Bai, 2015).

Genome-wide association study has been widely used in wheat (Breseghello and Sorrells, 2006; Juliana et al., 2015; Maccaferri et al., 2010, 2015; Zhang et al., 2011). The self-pollination nature of wheat results in a high level of LD, which reduces the minimum number of markers needed to detect marker-trait association. Several studies have examined the degree of LD in the wheat genome and found that LD, represented by squared allele frequency correlation coefficient (r2) value, decayed to <0.1 at about 10 to 20 cM (Chao et al., 2007; Zhang et al., 2010a; Juliana et al., 2015), <0.2 at 2 to 3 cM (Somers et al., 2007), and <0.3 at 1.6 cM (Maccaferri et al., 2015). These results suggested that GWAS can be performed with a limited number of molecular markers in wheat. Recently, GWAS has been successfully used to map gene and QTL for yield (Sukumaran et al., 2015), quality traits (Breseghello and Sorrells, 2006), leaf rust resistance (Maccaferri et al., 2010; Kertho et al., 2015; Gao et al., 2016), stripe rust resistance (Maccaferri et al., 2015; Kertho et al., 2015), stem rust resistance (Yu et al., 2012; Zhang et al., 2014), and soilborne mosaic virus resistance in wheat (Zhang et al., 2011).

The successful application of GWAS in wheat has prompted the development of genomic resources. A wheat HapMap has been constructed to facilitate GWAS in wheat (Jordan et al., 2015). Moreover, the USDA–ARS National Small Grains Collection (NSGC) has assembled a core subset of wheat accessions to maintain a high degree of genetic diversity with a manageable number of samples (Bonman et al., 2015). The NSGC wheat core subset consists of 4007 accessions, accounting for ∼10% of total common wheat germplasm in the NSGC. The core was assembled using the stratified random sampling approach based on the country of origin and may reserve rare genetic variations associated with traits of agronomic importance. Thus, the core subset of wheat accessions is a valuable resource for wheat improvement. The objective of this study was to identify novel leaf rust resistance genes in the core subset of wheat germplasm using GWAS and provide resistance sources for wheat breeding.

Materials and Methods

Plant Materials

The NSGC core set germplasm consists of 4007 accessions, and 3836 of them were genotyped with Illumina iSelect 9K SNP array and Diversity Arrays Technology (DArT) markers. Based on marker data, duplicate accessions were excluded from the core to form an informative core (iCore) with 3230 accessions including 1674 winter–facultative accessions and 1556 spring wheat accessions (Bonman et al., 2015). Of the 1674 winter–facultative accessions, 1596 were available at NSGC and used in this study. The details of these accessions, including name, origin, and improvement status, etc., are available at the TriticeaeCAP project website (http://www.triticeaecap.org/). This panel consists of 217 breeding lines, 589 cultivars, 773 landraces, and 17 genetic stocks and was collected from 85 countries with 594 from Asia, 794 from Europe, 15 from Africa, 114 from North America, 57 from South America, four from Central America, six from Australia, 10 from New Zealand, and two from the Caribbean islands. Experimental seeds were provided by the USDA–ARS NSGC. In addition, a set of 45 differential lines carrying known leaf rust resistance genes was also used in this study (Supplemental Table S1).

Evaluation of Leaf Rust Resistance

Three Pt pathotypes were used as inocula in this study including Pt2013, Pt52-2 (MMPSD), and Pt54-1 (TNBGJ). Pt2013 was a bulked Pt race collected in multiple fields of Oklahoma in 2013. Previous studies suggested that a bulk field collection of Pt pathotypes provided the broadest spectrum of virulence (Dr. Robert Hunger, personal communication, 2016), and leaf rust reaction to a mixture of races in the seedling test best resembled field reaction (Gao et al., 2016). Pt52-2 and Pt54-1 were two predominant races in Oklahoma fields and were collected and isolated by USDA–ARS Cereal Disease Laboratory at St. Paul, MN, in 2015.

All 1596 accessions were evaluated for Pt infection type (IT) in a greenhouse at the USDA–ARS Wheat, Peanut, and Other Field Crops Research Unit in Stillwater, OK. A randomized complete block design was used with two replicates. Eight seeds per entry were planted in each cell of a 73-cell growing tray (Growing Systems, Inc.) containing Sunshine Redi-earth Growing Mix (SunGro Horticulture Cabada Ltd.). Each tray had 70 cells for the test entries and three for the susceptible check, TAM110. Seedlings at the two-leaf stage (∼12 d) were inoculated with urediniospores of Pt2013 using a method described previously (Xing et al., 2014) in 2015. Inoculated seedlings were placed in a dew chamber with no lighting at ∼20°C with 100% relative humidity for 24 h and then transferred to a greenhouse at 22 ± 2°C with supplemental light and a photoperiod of 16:8 (light/dark) h. A 0-to-9 scale (McNeal, 1971) was used to score leaf rust IT of each plant when leaf rust was fully developed on susceptible checks at ∼14 d after inoculation. The mean IT of eight plants in each replicate was calculated for each accession and used for further analysis.

Based on the leaf rust IT data from 1596 accessions, a subset of accessions that met one of the following three criteria were chosen for a confirmation test in 2015: (i) an accession with a mean IT ≤ 3, (ii) plants with heterogeneous reactions to Pt2013 within a replicate, or (iii) the difference in IT ≥ 2 between two replicates. The purpose of the confirmation test was to get accurate data and identify reliable resistance sources, and the experimental design and disease evaluation protocols described above were employed in the confirmation test. The confirmation test identified 59 heterogeneous accessions, which were excluded from further analysis, leaving 1537 accessions available for association analysis. The leaf rust IT data of two replicates were highly consistent with a correlation coefficient of 0.947, and the mean IT over two replicates was calculated for each accession and used for GWAS analysis. In early 2016, the 1537 accessions were evaluated with the races Pt52-2 and Pt54-1 using the same experimental design.

Identification of the Lr47 Gene using a Diagnostic Marker

A total of 61 accessions showing high resistance to Pt2013 were tested for the presence or absence of the Lr47 gene using a T. speltoides (Tausch) Gren. ex K. Richt.-specific marker (Helguera et al., 2000). CItr 17884, a genetic stock carrying the Lr47 gene, was used as a positive control. Genomic DNA was extracted from seeds of each accession using a protocol reported by Dubcovsky et al. (1994). Primer sequences and polymerase chain reaction (PCR) conditions described previously were employed in this study (Helguera et al., 2000). The PCR products were separated by electrophoresis in 2% agarose gels and visualized using ethidium bromide and ultraviolet light.

Data Analysis

The method for collecting genotypic data of 1537 accessions was described previously (Bonman et al., 2015), and 5011 polymorphic SNP data from the TCAP project (http://www.triticeaecap.org/) were used for the GWAS.

Among the 5011 SNPs that were analyzed with HAPLOVIEW 4.2 (Barrett et al., 2005), 4091 were identified as nonredundant SNPs using the tagger function r2 = 1.0. A subset of 2110 informative SNPs were selected for structure (Q matrix) analysis using the tagger function r2 = 0.25. The Q matrix was estimated using the program STRUCTURE 2.3.4 (Pritchard et al., 2000) and was integrated as a covariate to correct the effects of population structure. An admixture model was used to test the hypotheses for two to 15 subpopulations (K = 2–15). For each K, five independent runs were performed with a burn-in length of 100,000, and the number of Markov Chain Monte Carlo iterations of 50,000. The log probability of data [LnP(D)] was estimated for each run, and an ad hoc statistic ΔK that was based on the rate of change in LnP(D) between successive K values was used to determine the true number of subpopulations (Evanno et al., 2005). Based on the inferred subpopulation number, a Q matrix representing population membership coefficients of each accession was obtained.

Of the 5011 SNPs, 4716 were mapped on a consensus reference map (Cavanagh et al., 2013; Wang et al., 2014) and thus used for genome-wide LD analysis. An R package (R Development Core Team, 2010) was used to calculate the LD for all pairwise comparisons between intrachromosomal SNPs and estimate the genome-wide LD decay by plotting LD r2 from all 21 chromosomes against the corresponding genetic distances. A trend line was fitted by second-degree loess (Cleveland 1979) using the statistical program R (R Development Core Team, 2010).

The mixed linear model (Yu et al., 2006) was employed to identify loci governing Pt response in the 1537 accessions using TASSEL 5.0 (Bradbury et al., 2007). The SNP markers and the Q matrix were used as fixed effects. To correct the family relatedness, a kinship (K) matrix was calculated with the scaled identity-by-state method that gives a reasonable estimate of additive genetic variance (Endelman and Jannink, 2012) and used as a random effect component in the MLM model. The MLM was run with the optimum level of compression and the P3D approach (Zhang et al., 2010b) using TASSEL 5.0. The FDR of 0.01 was used as a threshold to declare significant marker–trait association with the procedure developed by Benjamini and Hochberg (1995).


Linkage Disequilibrium Analysis

Among 4716 SNPs that were mapped in the 90K SNP consensus map (Cavanagh et al., 2013; Wang et al., 2014), 2220 were from the A genome, 2200 from the B genome, and 296 from the D genome. They were used to estimate genome-wide LD (Fig. 1). Based on the fitted model, genome-wide r2 values declined to 0.1 at 11 cM. Thus a genomic region covering 11 cM from each side of the peak of significant associations on the same chromosome arm was defined as a QTL-harboring region, and all significant associations in this region were considered as the same QTL.

Fig. 1.
Fig. 1.

Scatter plot showing the genome-wide linkage disequilibrium (LD) decay over genetic distances. Pair-wise LD r2 value is plotted against the intermarker map distance, which was based on a reference consensus map (Wang et al., 2014). The blue curve represents the fitted model, and the red dashed line shows LD declines to 0.1 at 11 cM.


Population Structure

The model-based structure analysis revealed six subpopulations (S1– S6), which include 339 (22.1%), 78 (5.1%), 437 (28.6%), 379 (24.7%), 133 (8.7%), and 171 (11.1%) accessions with the Fst values of 0.6502, 0.7943, 0.3240, 0.4588, 0.4323, and 0.4562 from S1 to S6, respectively (Fig. 2).

Fig. 2.
Fig. 2.

STRUCTURE analysis revealed six subpopulations (S1–S6). S1, S2, S4, S5, and S6 represent Iranian landrace group 1, eastern European cultivar group, western European cultivar group, east-Asian accession group, and Iranian landrace group 2, respectively. Origins of wheat accessions in S3 were very diverse. Vertical lines represent membership coefficients of accessions corresponding to the six subpopulations.


Iranian Landraces were predominant in S1, accounting for 87.6% in the subpopulation, and the remaining 12.4% originated from central Asia and Europe; therefore, this subpopulation was designated as Iranian landrace group 1. Iranian landraces were well represented in the iCore subset. Among 425 Iranian accessions used in this study, 297 belong to S1. Wheat cultivars from eastern Europe (Romania, Bulgaria, Russia, Croatia, and Ukraine) were predominant in S2, accounting for 87.2% (eastern European cultivar group). The others in this group were from Asia (9%), United States (2.6%), and South America (1.3%). The origins of accessions in S3 were diverse. The 437 accessions in this subpopulation, including 79 breeding lines, 207 registered cultivars, 139 landraces, and 12 genetic stocks, were collected from 60 countries in Europe, North America, Asia, Africa, South America, Australia, New Zealand, and the Caribbean islands (Cuba). European cultivars, especially western European cultivars, were predominant in S4, accounting for 77% of all accessions. The majority of cultivars released from western Europe, including Germany, England, the Netherlands, Sweden, Switzerland, Portugal, Spain, Italy, France, Finland, Demark, and Belgium, belonged to S4 (western European cultivar group). Some cultivars and landraces from other regions, including Asia, United States, and eastern Europe, were also included in this group. Subpopulation S5 was characterized by cultivars and landraces collected from eastern Asia, including China, Japan, and South Korea, and southern Asia including Bhutan, Nepal, and India. Accessions from these countries accounted for 68% of the total accessions in S5. A few landraces from western Asia (Iran), and one or two each from other 29 countries, were also included in this group (east-Asian accession group). Most (91.4%) of group S6 were landraces with a large portion (66.3%) from Iran. The remaining accessions in this group came from Turkey, Afghanistan, Macedonia, Armenia, Azerbaijan, Georgia, United States, Lebanon, and Jordan (Iranian landrace group 2).

Evaluation of Leaf Rust Reaction

Among the tested accessions, only a limited number of them (4%) showed a high level of seedling resistance to Pt2013 (highly resistant [HR]; IT = 0–3), and their mean IT distribution was significantly skewed toward susceptibility (Fig. 3). About 17.3, 36.6, and 43.1% were moderately resistant (MR; IT = 4–5), moderately susceptible (MS; IT = 6–7), and highly susceptible (HS; IT = 8–9), respectively. The paucity of HR germplasm was consistent with the fact that most known Lr genes have lost their effectiveness to Pt2013 (Supplemental Table S1). In the present study, we also tested a set of 45 differential lines with known Lr genes and found that only a few of them were highly resistant to Pt2013 (Supplemental Table S1) including lines with Lr19 (IT = 0;), Lr25 (IT = 1), Lr29 (IT = 3), Lr45 (IT = 0;), Lr47 (IT = 1), and Lr51 (IT = 3).

Fig. 3.
Fig. 3.

Distribution of mean leaf rust infection types of 1537 accessions inoculated with Pt2013, Pt52-2, and Pt54-1.


Compared with Pt2013, Pt52-2 and Pt54-1 were less virulent. About 8.0, 4.0, 18.6, and 69.4% accessions in the association panel was rated as HR, MR, MS, and HS to race Pt52-2, respectively, and 7.4, 12.2, 36.1, and 44.3% as HR, MR, MS, and HS to Pt54-1, respectively. These results were in agreement with the leaf rust reactions of differential lines carrying known resistance genes. Pt52-2 was virulent to Lr3, Lr3bg, Lr3ka, Lr10, Lr12, Lr14a, Lr14b, Lr15, Lr17, Lr20, Lr22a, Lr35, Lr37, Lr13, and Lr23 and avirulent to Lr2a, Lr2b, Lr2c, Lr18, Lr19, Lr21, Lr24, Lr25, Lr28, Lr29, Lr36, Lr38, Lr45, Lr47, Lr51, Lr52, Lr60, Lr63, and Lr64. Isolate Pt54-1 was highly virulent to Lr10 and avirulent to Lr18, Lr19, Lr21, Lr25, Lr29, Lr36, Lr38, Lr45, Lr47, Lr51, Lr52, Lr60, Lr63, and Lr64 (Supplemental Table S1). Other genes confer MR or MS reactions to Pt54-1.

Significant association was observed between population structure and leaf rust resistance, suggesting that population structure should be considered in GWAS. For example, ANOVA among subpopulations revealed significant variation in IT scores for Pt2013 (p < 0.01) (Fig. 4). The subpopulation S2 showed the highest level of resistance as a group with a mean IT score of 4.04, which was significantly lower than all other five groups (p < 0.01). Of the 78 accessions in this group, the majority were rated HR (34.6%) or MR (35.9%), and only a few of them were rated MS (16.7%) or HS (12.8%). In contrast, none or only a few HR accessions were identified in the subpopulations S1 and S6, the two Iranian landrace groups. In the S1 group, 0, 2.1, 30.1, and 67.8% of the accessions were rated HR, MR, MS, and HS, respectively, and in the S6 group, these numbers were 0.6, 9.4, 32.7, and 57.3%, respectively. The mean IT scores of S3 (6.53) and S5 (6.83) were significantly lower than those of S1 (7.61) and S6 (7.34) (p < 0.01) but significantly higher than that of S4 (5.91) (p < 0.01). Similar results were observed when the association panel was inoculated with Pt52-2 and Pt54-1.

Fig. 4.
Fig. 4.

Box plot showing distribution of wheat leaf rust infection types (ITs) in six subpopulations inoculated with Pt2013. The diamond represents mean IT value of each subpopulation.


Quantitative Trait Loci for Leaf Rust Resistance

The MLM analysis revealed a total of 14 QTL that were significantly associated with leaf rust resistance at FDR of 0.01 in three experiments (Table 1). Six, eight, and five significant QTL were detected for resistance to the pathotypes Pt2013, Pt52-2, and Pt54-1, respectively. Each QTL explained 1.36 to 8.35% of the phenotypic variance for resistance to Pt2013, 1.35 to 7.23% for resistance to Pt52-2, and 1.48 to 2.64% for resistance to Pt54-1. Of these QTL, QLr.stars-1BC1 was significant for resistance to all three pathotypes, while QLr.stars-1BS1 and QLr.stars-7AL1 were significant for resistance to pathotypes Pt2013 and Pt54-1. QLr.stars-2DS1 was significant for resistance to Pt52-2 and Pt54-1. The remaining 10 QTL were significant for resistance to only a single pathotype at FDR of 0.01 (Table 1). Among them, however, QLr.stars-1DC1, QLr.stars-1BL1, QLr.stars-1BL2, QLr.stars-4AL1, and QLr.stars-4BL1 were significant for resistance to two to three pathotypes if a FDR was set at 0.05 (Supplemental Table S1). Moreover, 31 additional significant loci were detected at a FDR of 0.05 (Supplemental Table S1).

View Full Table | Close Full ViewTable 1.

Designated name, representative single-nucleotide polymorphism (SNP; TagSNP), chromosome arm, location on the Illumina SNP9K consensus map, p-value, and R2-value of each quantitative trait loci (QTL) that was significant at a false discovery rate of 0.01 when the association panel was evaluated using Pt pathotypes Pt2013, Pt52-2, and Pt54-1.

Pt pathotype QTL TagSNP SNP Chromosome arm† Map location p-value R2
cM %
Pt2013 QLr.stars-1BC1 IWA435 5 1BC 60.13 2.09 × 10−13 4.76
QLr.stars-1BL1 IWA8153 1 1BL 148.23 3.31 × 10−5 1.36
QLr.stars-1BS1 IWA1566 7 1BS 48.08–57.72 2.31 × 10−11 3.64
QLr.stars-1BS1 IWA6110 3 1BS 9.06 × 10−15 4.83
QLr.stars-1DC1 IWA362 1 1DC 48.81 1.73 × 10−5 1.84
QLr.stars-2DL1 IWA1083 14 2DL 176.19–177.99 1.67 × 10−6 1.82
QLr.stars-7AL1 IWA6736 3 7AL 188.95 5.71 × 10−24 8.35
Pt52-2 QLr.stars-1BC1 IWA435 1 1BC 60.13 2.25 × 10−5 1.43
QLr.stars-1BL2 IWA6758 1 1BL 81.58 5.96 × 10−6 1.60
QLr.stars-1BL3 IWA579 3 1BL 115.79 4.08 × 10−5 1.35
QLr.stars-2DS1 IWA609 6 2DS 79.50–92.80 2.19 × 10−23 7.23
QLr.stars-4AL1 IWA2460 2 4AL 73.40–83.38 2.65 × 10−5 1.40
QLr.stars-4BL1 IWA6480 2 4BL 72.30–82.13 2.11 × 10−6 1.75
QLr.stars-5BL1 IWA4571 4 5BL 106.17–106.94 1.85 × 10−6 1.75
QLr.stars-7AS1 IWA3760 1 7AS 45.42 6.38 × 10−6 1.59
Pt54-1 QLr.stars-7AL1 IWA6736 1 7AL 188.95 6.29 × 10−8 1.95
QLr.stars-1BC1 IWA435 1 1BS 42.93 3.43 × 10−7 1.97
QLr.stars-2DS1 IWA4012 3 2DS 92.80–93.40 7.87 × 10−7 2.64
QLr.stars-2BL1 IWA1488 1 2BL 193.90 1.31 × 10−5 1.48
QLr.stars-1BS2 IWA6110 3 1BS 1.87 × 10−10 1.79
L, S, and C refer to long and short chromosome arms, and centromere, respectively.

The 14 QTL significant at FDR of 0.01 were located on eight chromosomes including five loci on chromosome 1B, two each on chromosomes 2D and 7A, and one each on chromosomes 1D, 2B, 4A, 4B, and 5B (Table 1). For two of the QTL on 1B, QLr.stars-1BS1 is on the short arm, while QLr.stars-1BC1 is in the centromere region where recombination is usually suppressed. The genetic distance between them is <10 cM. However, the physical distance between them may be greater than that indicated by the genetic distance because of recombination suppression. We further calculated the LD between the two tag SNPs underlying these two QTL, IWA1566 and IWA435, and found that the LD was only 0.089, suggesting that these two QTL were likely different loci.

Three SNPs cosegregating in the association panel, IWA6110, IWA6541 and IWA6129, were significantly associated with leaf rust resistance to both pathotypes Pt2013 and Pt54-1. These SNPs have not been mapped to the 90K SNP consensus map yet. To determine the genomic locations of IWA6110, IWA6129, and IWA6541, we compared their sequences (∼200 bp) with other SNP sequences from 90K SNP arrays and found that IWA6110 and IWA6541 were identical to IWB3972 and IWB63777, respectively. IWB3972 was previously mapped to the centromere region of chromosome 1B and IWB63777 on the proximal region of 1BS. IWB3792 and IWB6377 were 1.54 cM apart on the 90K consensus map (Fig. 5). We further searched these three SNP sequences against Chinese Spring reference genome sequence (urgi.versailles.inra.fr/blast/blast.php). Three contigs mapped to chromosome 1B, including 1BS_3441348, 1BS_3445781, and 1BS_3456976, contained IWA6541, IWA6129, and IWA6110 sequence, respectively. We searched these contigs against Brachypodium physical map using BLASTn (http://www.plantgdb.org/BdGDB/cgi-bin/blastGDB.pl) and located them to 23.8, 34.8, and 59.8 Mb of chromosome 2, respectively (Fig. 5). Considering the collinearity between wheat and the Brachypodium genome (Brenchley et al., 2012), we located IWA6129 in between IWA6110 and IWA6541 at the 58.37 to 59.91 cM region on the consensus map. Given that LDs between these SNPs and two tag SNPs, IWA1566 and IWA435, were estimated to be 0.2 and 0.06, respectively, we conclude that these SNPs represent the same locus as IWA1566, QLr.stars-1BS1.

Fig. 5.
Fig. 5.

Inferred locations of single-nucleotide polymorphism (SNP) marker IWA6541, IWA6110, and IWA6129 on the SNP9K consensus map (right) and homeologous sequence corresponding to each marker on the Brachypodium chromosome 2 physical map (left). IWA6129 is tentatively mapped in between IWA6541 and IWA6110.


Three QTL were detected on 1BL. QLr.stars-1BL1 was significant for resistance to Pt2013, while QLr.stars-1BL2 and QLr.stars-1BL3 were significant for resistance to Pt52-2. Their locations on the 90K consensus map are 148.23, 81.58, and 115.79 cM, respectively, thus they are likely different QTL. The other nine QTL were identified in different chromosome arms of seven chromosomes (Fig. 6).

Fig. 6.
Fig. 6.

Chromosome locations of 14 quantitative trait loci (QTL) for leaf rust resistance identified in this study and previously mapped Lr genes and QTL (left side of bars). The vertical lines on the left show genomic regions harboring known genes or QTL. Chromosome lengths were standardized to the same relative length. Marker positions were based on the 90K SNP consensus linkage map. The tag SNP of each QTL was shown on the linkage map, and suggested QTL name was given in parentheses.


Allele Effects of the Quantitative Trait Loci Identified in This Study

In the Pt2013 experiment, the allelic effects of the six QTL identified in this study varied from 0.26 to 2.27 for IT score. For two of them, allele substitution resulted in IT change >1.0 (Table 2). The favorable allele number in each accession ranged from 0 to 5 with an average of 1.57 at six loci. The mean IT values decreased from 7.42 for germplasm without any favorable allele to 3.82 for those possessing five favorable alleles (Supplemental Fig. S1). All 23 accessions having five favorable alleles were rated as either HR or MR to Pt2013. It is of great interest to know whether the leaf rust resistance of these accessions would prove to be durable.

View Full Table | Close Full ViewTable 2.

Allelic effects of the tag single-nucleotide polymorphisms (SNPs) underlying quantitative trait loci (QTL) for leaf rust resistance and the distribution of favorable alleles in six subpopulations

QTL name TagSNP SNP allele† Allele effects
Favorable allele frequency‡
Pt2013 Pt52-2 Pt54-1 S1 S2 S3 S4 S5 S6 All
————————————————— % —————————————————
QLr.stars-1BC1 IWA435 T/C 0.72 0.80 0.37 0.9 6.4 17.6 3.4 29.3 8.2 9.8
QLr.stars-1BL1 IWA8153 T/C 0.52 2.1 30.8 29.3 66.0 15.0 11.7 29.2
QLr.stars-1BS1 IWA6110 A/G 2.27 1.70 0.0 25.6 0.5 5.5 0.0 0.0 2.8
QLr.stars-1DC1 IWA362 T/C 0.26 78.8 16.7 35.2 29.3 77.4 56.1 48.4
QLr.stars-2DL1 IWA1083 T/C 0.60 2.1 30.8 29.7 65.4 14.3 11.7 29.1
QLr.stars-7AL1 IWA6736 T/C 1.26 0.69 5.0 74.4 33.4 67.8 6.8 19.9 33.9
QLr.stars-1BL2 IWA6758 A/G 0.40 2.1 6.4 14.9 10.3 44.4 15.2 11.6
QLr.stars-1BL3 IWA579 T/C 0.31 40.6 6.4 9.8 9.2 84.2 68.4 29.3
QLr.stars-2BL1 IWA1488 A/G 1.00 3.8 0.0 1.6 4.2 14.3 7.0 4.4
QLr.stars-2DS1 IWA609 A/G 2.50 3.2 5.1 2.3 2.1 23.8 9.0 5.1
QLr.stars-2DS1 IWA4012 A/G 1.02 5.3 22.2 4.5 7.7 29.2 14.9 10.2
QLr.stars-4AL1 IWA2460 T/C 0.42 1.5 3.8 13.7 8.4 70.7 14.6 14.2
QLr.stars-4BL1 IWA6480 A/C 0.56 96.5 16.7 28.4 20.3 78.9 97.7 52.0
QLr.stars-5BL1 IWA4571 T/C 0.30 81.4 88.5 71.6 92.3 70.7 93.0 80.9
QLr.stars-7AS1 IWA3760 A/C 0.48 1.8 74.4 33.4 71.0 14.3 36.3 39.2
The favorable alleles are in bold type.
S1, Iranian landrace group 1; S2, eastern European cultivar group; S4, western European cultivar group; S5, east-Asian accession group; S6, Iranian landrace group 2.

A significant correlation between favorable allele numbers and IT values was observed (r = −0.401, p < 10−7). The observed phenotypic values and the predicated phenotypic values (calculated as the sum of all QTL effects for each accession) were also significantly correlated (r = −0.53, p < 10−7), suggesting that pyramiding these QTL may enhance leaf rust resistance.

Similarly, favorable alleles were negatively correlated with the IT scores when plants were inoculated with Pt52-2 (r = −0.35, p < 10−7) and Pt54-1 (r = −0.45, p < 10−7). The favorable alleles in each accession varied from 0 to 7 with a mean allele number of 2.4 at eight QTL resistant to Pt52-2, and 0 to 4 with an average of 0.6 at five loci resistant to Pt54-1. The allele substitution effects for IT score ranged from 0.3 to 2.5 and 0.37 to 1.7 in the two experiments, respectively (Table 2). Notably, the favorable allele of QLr.stars-1BS1 was not detected in east-Asian cultivars or landraces (S5), while the favorable allele of QLr.stars-2BL1 was not present in cultivars from eastern Europe (S2)(Table 2), suggesting that the two QTL may diversify leaf rust resistance sources for breeding programs in eastern Asia and eastern Europe, respectively.


The availability of high-density consensus linkage maps makes it feasible to determine the relationship of QTL identified in this study with known Lr genes or QTL. Based on 90K SNP consensus map (Wang et al., 2014), the relationship between the known loci for leaf rust resistance and the genomic regions harboring QTL identified in this study was illustrated in Fig. 6.

Two named genes, Lr71 and LrZh84, and three QTL, QLr.pser-1BL, QLr.sfr-1BS, and QLr.sfr-1B, for leaf rust resistance were previously mapped to the genomic regions harboring QLr.stars-1BC1 and QLr.stars-1BS1 (Zhou et al., 2013; Li and Bai, 2009; Schnurbusch et al., 2004; Messmer et al., 2000; Singh et al., 2013). Lr71 is a gene identified in spelt wheat cultivar Altgold Rotkorn (Singh et al., 2013). Given that no spelt wheat was included in the association panel, QLr.stars-1BC1 and QLr.stars-1BS1 are unlikely Lr71. LrZh84, likely originated from wheat cultivar Predgomaia, has been effective in the field for >30 yr in the Huang-Huai Valley winter wheat zone of China (Zhao et al., 2008; Zhou et al., 2013), while both QLr.sfr-1BS and QLr.sfr-1B express slow leaf rusting resistance in wheat cultivar Forno (Messmer et al., 2000; Schnurbusch et al., 2004). Another QTL mapped in this region, QLr.pser-1BL, showed adult plant resistance (Li and Bai, 2009). Thus, the relationship between these known genes or QTL and the QTL identified in this study remain to be investigated.

Similarly, QLr.stars-1BL1, QLr.stars-1BL2, QLr.stars-2DS1, QLr.stars-2DL1, QLr.stars-4AL1, and QLr.stars-4BL1 are in the proximity of known genes or QTL (Fig. 6). QLr.stars-1BL1 and QLr.stars-1BL2 reside in the genomic regions where Lr46 and Lr33 were located, respectively (William et al., 2003; Dyck and Sykes, 1994). Lr46 is an important durable leaf rust resistance gene that exhibits adult plant resistance. Lr46 was first identified in the wheat cultivar Pavon. The effect of Lr46 resembles that of Lr34 at the adult plant stage but is weaker than that of Lr34 at the seedling stage (Martínez et al., 2001). The QLr.stars-1BL1 was significant at FDR0.01 in the Pt52-1 experiment and FDR0.05 in the Pt2013 experiment. The allele effect of this QTL was relatively small (Table 2), suggesting that, with accurate phenotypic data, genes with minor effects can be identified using GWAS. QLr.stars-1BL1 and QLr.stars-1BL2 may represent Lr46 and Lr33, respectively. QLr.stars-2DL1 was in the vicinity of Lr54 and a QTL identified in the wheat cultivar Arina (Schnurbusch et al., 2004) (Fig. 6). Lr54 originated from chromosome 2L of Ae. kotschyi (Marais et al., 2005; Heyns et al., 2011). Given that genetic stocks known to carry Lr54 were not included in this association panel, QLr.stars-2D may be associated with the QTL identified in wheat cultivar Arina, rather than Lr54. In addition, QLr.stars-2DS1, QLr.stars-4AL1, and QLr.stars-4BL1 coincided with Lr2 (Lr15), Lr30, and Lr12 (Lr31) (Singh and Bowden, 2011), respectively. Further studies may facilitate determination of the relationship between known resistance genes and QTL identified in this study.

QLr.stars-7AS1 is <4 cM away from Lr47 (Fig. 6). Lr47, transferred from chromosome 7S of T. speltoides to chromosome 7A of common wheat (Dubcovsky et al., 1998), was highly resistant to all three pathotypes used in this study (Supplemental Table S1). To determine the relationship between QLr.stars-7AS1 and Lr47, we used a diagnostic marker (Helguera et al., 2000) to detect wheat accessions carrying the Lr47 gene from those showing high resistance to Pt2013. Our results suggested that no accession in this panel carries Lr47. Thus QLr.stars-7AS1 is a novel locus.

QLr.stars-7AL1 was significantly associated with resistance to Pt2013 and Pt54-1 and is 15.4 cM distal to Lr20 on the linkage map (Neu et al., 2002; Wang et al., 2014), indicating that QLr.stars-7AL1 may be independent from Lr20 and likely a novel locus in bread wheat. A recent GWAS study revealed that Xcfa2257, a simple-sequence repeat marker that is 2 cM distal to QLr.stars-7AL1, was associated with leaf rust resistance in durum wheat (Maccaferri et al., 2010), lending credence to QLr.stars-7AL1. The SNP allele associated with the resistant form of this QTL was present at high frequencies in the eastern European cultivar group (74.4%) and the western European cultivar group (67.8%) but at a low frequency in the east-Asian accession group (6.8%). Therefore, it is worthy to evaluate the effectiveness of this QTL against Pt races in eastern Asia, and it may diversify leaf rust resistance sources in this region if it turns out to be resistant.

Another four QTL, QLr.stars-1BL3, QLr.stars-1DC1, QLr.stars-2BL1, and QLr.stars-5BL1, are likely novel loci for leaf rust resistance. Based on locations of molecular markers flanking known Lr genes on the 90K consensus map, QLr.stars-1BL3 is 30 cM from QLr.stars-1BL2/Lr33 and at least 17.5 cM from Lr46 (William et al., 2003). The genetic distance between QLr.stars-2BL1 and the two closest leaf rust resistance genes, Lr35 and Lr13 (Seyfarth et al., 1999, 2000), is at least 27.2 cM (Fig. 6). Moreover, no gene has been identified in the genomic regions harboring QLr.stars-1DC1 and QLr.stars-5BL1. These novel loci represent valuable additions to the leaf rust resistance gene pool available for wheat breeding.

QTL identified in this study showed obvious regional distribution. For example, 10 SNPs at the QLr.stars-1BS1 locus were significantly associated with leaf rust resistance, and three of them, IWA6110, IWA6129, and IWA6541, showed the largest effect. The favorable alleles of these SNPs were mainly found in eastern and western European cultivars (S2 and S4), and were completely absent in east-Asian accessions and Iranian landraces (S1, S4, and S5). Thus, if QLr.stars-1BS1 is effective to Pt races in eastern Asia, it should be a valuable resistance source for wheat breeding programs in this region. On the contrary, the favorable allele of IWA1488, the tagSNP of QLr.stars-2BL1, is present at a relatively higher frequency in east-Asian accessions (S3), and no cultivars in eastern Europe (S2) carry this allele. Use of this QTL may diversify resistance sources if it confers resistance to Pt races in this region. Given that SNP markers associated with these QTL are available and high-throughput assays, such as Kompetitive Allele-Specific PCR, can be easily developed for these markers (Cabral et al., 2014), QTL identified in this study can be reliably introgressed into elite cultivars using marker-assisted selection.


Genome-wide association study is an effective gene discovery approach complementary to the traditional linkage mapping. In this study, 14 QTL were significantly associated with leaf rust resistance at a FDR of 0.01. Of these, eight QTL were mapped in the vicinity of known genes or QTL in bread wheat. Allelism tests may confirm their relationship with known genes or QTL. QLr.stars-7AL1 was the most significant QTL identified in this study, explaining 8.4% of the total phenotypic variance. This QTL is new to bread wheat but is close to a locus previously identified in durum wheat. The other five QTL, including QLr.stars-1BL3, QLr.stars-1DC1, QLr.stars-2BL1, QLr.stars-5BL1, and QLr.stars-7AS1, are likely novel loci for leaf rust resistance. The distribution of these 14 QTL was uneven in the six subpopulations, suggesting that wheat breeding programs can enhance leaf rust resistance by introgressing specific QTL identified in this study into new cultivars. In addition, another 31 loci were significantly associated with leaf rust resistance at a less stringent FDR of 0.05 and will be valuable additions to the leaf rust resistance gene pool.

Supplemental Information Available

Supplemental information is available with the online version of this manuscript.


We thank Dr. Harold Bockelman of USDA–ARS National Small Grain Collection for providing germplasm used in this study, the National Research Initiative Competitive Grant 2011-68002-30029 (Triticeae-CAP) from the USDA National Institute of Food and Agriculture for providing SNP data, and Todd Lenger for his excellent technical support. Mention of trade names or commercial products in this article is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the USDA. The USDA is equal-opportunity provider and employer.





Be the first to comment.

Please log in to post a comment.
*Society members, certified professionals, and authors are permitted to comment.