Breeding for wheat stripe rust: understanding genetic architecture and evaluating genomic prediction in the Uruguayan breeding program

Abstract

Wheat yellow (stripe) rust, caused by Puccinia striiformis f. sp. tritici (Pst), is one of the most devastating diseases of wheat worldwide. Yellow rust was a disease considered marginal in Uruguay until 2016. Since the severe epidemic in 2017, yellow rust has become the most economically important wheat foliar disease in Uruguay. In the short term, local management of ongoing epidemics on susceptible wheat varieties is limited to fungicide spraying. However, the deployment of resistant wheat cultivars appears as the most environmentally friendly strategy without additional cost for producers. Until 2017, the Uruguayan Wheat Breeding Program (UWBP) did not consider resistance to Pst as a priority, therefore little is known about the genetic architecture of the disease resistance in local genetic germplasm, a key requisite for developing resistant varieties. With the general objective of contributing to the increase and sustainability of national wheat production through the development of yellow rust resistant cultivars, we defined a diverse mapping population to 1) characterize wheat resistance architecture to yellow rust using genome-wide association studies (GWAS) and 2) predict wheat lines behavior to yellow rust employing genomic prediction (GP) models. The GWAS population consisted of 368 wheat lines, mostly from the UWBP and the main cultivars grown locally. The population was phenotyped in the field for two consecutive years (2021 and 2022) under artificial inoculations. The response variable used was the area under the disease progress curve (AUDPC) based on six independent evaluation dates each year. To obtain the adjusted means for each year, a model based on the field trial of incomplete resolvable block was compared with models that include spatial adjustments as row and column effect and/or spatial correlations between residuals. Then, adjusted means between years were obtained because to correlation between the means obtained by year was high (0.84). Additionally, the population was genotyped using genotyping by sequencing (GBS), the SNPs were called using TASSEL 5 and imputation of missing data was performed with Beagle. Over 100 thousand SNPs were obtained after filtering. GWAS analyses were performed with GAPIT R package. Genomic regions associated with resistance to Pst identified on chromosomes 1A, 1B, 2B, 3A, 5A, 5B, 5D, 7A, and 7B were significant in at least two of the implemented models. Next step will be to search the available databases to identify if any gene or quantitative trait loci (QTL) associated with resistance to the disease was reported in the identified regions. For GP, the correlation between adjusted phenotypic means and predicted values (prediction accuracy) was calculated using 10-fold cross-validation with the GBLUP model in the rrBLUP package of R Software. The prediction accuracy obtained was on average 0.64, which positions the GP as a promising tool for the selection of resistant lines in the UWBP.