BayesR and MLM association mapping approaches in common wheat landraces were used to identify genomic regions conferring resistance to Yr, Lr, and Sr diseases. Deployment of rust resistant cultivars is the most economically effective and environmentally friendly strategy to control rust diseases in wheat. However, the highly evolving nature of wheat rust pathogens demands continued identification, characterization, and transfer of new resistance alleles into new varieties to achieve durable rust control. In this study, we undertook genome-wide association studies (GWAS) using a mixed linear model (MLM) and the Bayesian multilocus method (BayesR) to identify QTL contributing to leaf rust (Lr), stem rust (Sr), and stripe rust (Yr) resistance. Our study included 676 pre-Green Revolution common wheat landrace accessions collected in the 1920-1930s by A.E. Watkins. We show that both methods produce similar results, although BayesR had reduced background signals, enabling clearer definition of QTL positions. For the three rust diseases, we found 5 (Lr), 14 (Yr), and 11 (Sr) SNPs significant in both methods above stringent false-discovery rate thresholds. Validation of marker-trait associations with known rust QTL from the literature and additional genotypic and phenotypic characterisation of biparental populations showed that the landraces harbour both previously mapped and potentially new genes for resistance to rust diseases. Our results demonstrate that pre-Green Revolution landraces provide a rich source of genes to increase genetic diversity for rust resistance to facilitate the development of wheat varieties with more durable rust resistance.

No abstract

The quantitative trait loci QYr.sun-3BS and QYr.sun-4DL were identified in the W195/BTSS recombinant inbred line (RIL) population in a previous study. QYr.sun-3BS explained 34 to 59% phenotypic variation in stripe rust response. We evaluated parental genotypes at different growth stages and temperature regimes to detect the critical stage for expression of QYr.sun-3BS. W195 expressed low infection type (IT) ;1C at the fourth leaf stage, when incubated at 21 ± 2°C and the alternate parent BTSS was susceptible (IT 3). Monogenic segregation for stripe rust response was observed among the RIL population at the fourth leaf stage and the underlying locus was temporarily named YrW195. YrW195 corresponded to QYr.sun-3BS. Since no previously designated stripe rust resistance genes that expresses at and after the fourth leaf stage was mapped in this region, YrW195 was formally named Yr58. Genotyping with Yr46-linked markers indicated the presence of Yr46 in W195, which corresponded to QYr.sun-4DL. The RILs carrying Yr58 and Yr46 singly produced IT 23C and IT 3, respectively, and those carrying both genes produced IT ;1C indicating the enhancement of Yr58 expression by Yr46. The absence of Yr58-linked alleles of markers sun533 and sun476 in 74 of the 76 wheat cultivars demonstrated their usefulness for marker-assisted selection.

Global wheat production is constrained by rust diseases and deployment of combinations of genetically diverse sources of resistance in new cultivars is accepted as a preferred means of rust control. This investigation was planned to identify new genomic regions contributing towards resistance to three rust diseases under field conditions in a W195/ BT-Schomburgk Selection (BTSS) RIL population. The RIL population was assessed for variation in stripe rust, leaf rust and stem rust responses at two locations during 2012 and 2013 crop seasons and was subjected to DArTseq genotyping. Linkage map was constructed using 3439 DArTseq markers with an average marker density of 2.7 cM. Composite interval mapping detected three QTL each for stripe rust, leaf rust and stem rust resistance in W195/BTSS RIL population. Two consistent QTL for stripe rust, QYr.sun-3BS and QYr.sun-4DL, were contributed by W195 and one inconsistent QTL, QYr.sun-7AS, by BTSS. In the case of leaf rust QLr.sun-2BS and QLr.sun-3BS were contributed by BTSS and QLr.sun-4DL by W195. Based on seedling tests, QLr.sun-2BS was demonstrated to be Lr23. BTSS carried stem rust QTL; QSr.sun-2BL and QSr.sun-6AS and QSr.sun-4DL was contributed by W195. QSr.sun-6AS corresponded to stem rust resistance gene Sr8a carried by BTSS. The collocated QTL for all three rust diseases in chromosome 4D corresponded to the plieotropic locus Lr67/Yr46/Sr55/Pm46 based on genotyping with a linked SNP marker. Interaction among QTL to condition lower rust responses was also demonstrated. QYr.sun-3BS, QLr.sun-3BS and QSr.sun-2BL represent new rust resistance loci.

Genomic selection can increase the rate of genetic gain in crops through accumulation of positive alleles and reduce phenotyping costs by shortening the breeding cycle time. We performed genomic prediction for resistance to wheat rusts in tetraploid wheat accessions using three cross-validation with the objective of predicting: (1) rust resistance when individuals are not tested in all environments/locations, (2) the performance of lines across years, and (3) adult plant resistance (APR) of lines with bivariate models. The rationale for the latter is that seedling assays are faster and could increase prediction accuracy for APR. Predictions were derived from adult plant and seedling responses for leaf rust (Lr), stem rust (Sr) and stripe rust (Yr) in a panel of 391 accessions grown across multiple years and locations and genotyped using 16,483 single nucleotide polymorphisms. Different Bayesian models and genomic best linear unbiased prediction yielded similar accuracies for all traits. Site and year prediction accuracies for Lr and Yr ranged between 0.56–0.71 for Lr and 0.51–0.56 for Yr. While prediction accuracy for Sr was variable across different sites, accuracies for Yr were similar across different years and sites. The changes in accuracies can reflect higher genotype × environment (G × E) interactions due to climate or pathogenic variation. The use of seedling assays in genomic prediction was underscored by significant positive genetic correlations between all stage resistance (ASR) and APR (Lr: 0.45, Sr: 0.65, Yr: 0.50). Incorporating seedling phenotypes in the bivariate genomic approach increased prediction accuracy for all three rust diseases. Our work suggests that the underlying plant-host response to pathogens in the field and greenhouse screens is genetically correlated, but likely highly polygenic and therefore difficult to detect at the individual gene level. Overall, genomic prediction accuracies were in the range suitable for selection in early generations of the breeding cycle.