Molecular markers associated with barley yellow dwarf virus tolerance in spring oat and their utilization in predictive breeding

Barley yellow dwarf viruses (BYDVs) are responsible for the disease barley yellow dwarf (BYD), which causes significant yield losses in many cereals including oat (Avena sativa L.). Phenotyping for disease sensitivity is time consuming, laborious and requires viruliferous aphids for inoculations. Until recently, the molecular marker technology in oat has not allowed for many marker-trait association studies to determine the genetic mechanisms for tolerance and as a result, marker-assisted selection (MAS) and genomic selection (GS) have not been extensively used in breeding for BYD tolerance. In the first study, a genome-wide association study (GWAS) was performed on 428 spring oat lines using a recently developed high-density oat single nucleotide polymorphism (SNP) array as well as a SNP-based consensus map. Marker-trait associations were performed using a Q-K mixed model approach to control for population structure and relatedness. Six significant SNP-trait associations representing two QTL were found on chromosomes 3C and 18D. This is the first report of BYDV tolerance QTL on chromosome 3C and 18D. Haplotypes using the two QTL were evaluated, and distinct classes for tolerance were identified based on the number of favorable alleles. In the second study, GS and MAS models were compared in their accuracy to predict barley yellow dwarf virus tolerance in 428 spring oat lines from North America and Europe. Several GS models were evaluated using 2305 SNPs and included models with previously identified or randomly selected markers as fixed effects. Model accuracies were evaluated using five-fold cross evaluation. GS models used ridge regression-best linear unbiased predictor (RR-BLUP) for marker effect estimation while MAS models used ordinary least square (OLS). Moderate to high prediction accuracies (0.5-0.9) were observed across the models. GS models containing fixed effects (GS-GWAS, GS-3C18D) from previously identified QTL performed better than the GS model with all markers as random effects or the MAS models. Two MAS models (MAS-GWAS and MAS-3C18D) had prediction accuracies higher than the GS model with all markers as random effects. In the third study, GS was used to identify individuals with high BYDV tolerance for use in cross prediction. To do this, 2138 SNPs were used on a panel of 519 spring oat lines for barley yellow dwarf virus tolerance. Of the 519 oat lines, 428 lines had genotypes and phenotypes while 91 of the oat lines were only genotyped. Using the R package “PopVar”, several GS models were compared for prediction accuracy. The BayesA model was identified as having the highest prediction accuracy and genomic estimated breeding values (GEBVs) were calculated for the 519 lines using the BayesA model. The top 10% of lines (52 lines) based on GEBVs for BYDV tolerance were selected to perform simulated crosses. A total of 1326 crosses were simulated, and the mean, genetic variance and mean of high/low superior progenies were calculated. From the 1326 crosses, 22 crosses were identified as having a balance between a low predicted mean (high tolerance) and high genetic variance. Because of the high tolerance and high genetic variance, the chance of obtaining transgressive segregants for BYDV tolerance is higher in these crosses. Using GS and simulated crosses gives breeders additional tools to improve breeding efficiency for BYDV tolerance and allows for better allocation of time and resources within the breeding program