Breeding hard winter wheat (Triticum aestivum L.) for high grain yield and high grain protein concentration

2021 Spring.Includes bibliographical references.High grain yield (GY) is the primary selection target in commercial hard winter wheat (Triticum aestivum L.) breeding programs, with milling and bread-making quality as important secondary selection targets. Grain protein concentration (GPRO) is strongly correlated with important dough rheology and bread-making characteristics. Simultaneous improvement is difficult given the strong negative relationship of GY and GPRO in cereal crops. Nitrogen use efficiency (NUE), defined as the amount of grain produced per unit of N supply, promotes high GY through the component traits N uptake (NUpE) and N utilization (NUtE) efficiencies. Grain protein accumulation relies on N uptake from the soil and remobilization from plant tissue reserves. One study was conducted to characterize variation for NUE among a set of 20 breeding lines and varieties adapted to the west central Great Plains of the United States. Path analysis was applied to characterize the NUE component structure during the 2010-2011 growing season and then for two newly released varieties in the 2011-2012 growing season. Nitrogen use efficiency ranged from 39.9 kg kg-1 for 'RonL' to 46.7 kg kg-1 for 'Byrd'. By path analysis, we determined that variation in NUE depended on NUpE under N sufficiency and on NUtE under limiting N. Additionally, strategies for simultaneous improvement of GY and GPRO were explored. Analysis of standardized residuals of the linear regression of GPRO on GY, or 'grain protein deviation', identified one cultivar ('Brawl CL Plus') that had 6.7 g kg-1 higher GPRO than the average for all 20 genotypes. In a second study, selection strategies based on protein-yield selection indices for a set of 775 breeding lines and varieties representing the Colorado State University hard winter wheat breeding program were evaluated based on field data obtained during the 2012-2015 growing seasons. Selection based on high values for a particular index delivered a characteristic emphasis on GY or GPRO. Correlation analysis between index values and GY or GPRO showed that each simultaneous selection strategy focused to differing extents on the primary traits. Genomic selection applied to index values in univariate models provided forward prediction accuracy ranging between r = .21 to .44 for the 2013 validation set, but approached zero for the 2014 validation set. Index values were also calculated from genomic estimated breeding values obtained in bivariate genomic selection models. Prediction accuracy for individual trait values was not substantially improved in the bivariate model. Protein-yield indices calculated from bivariate genomic estimated breeding values showed similar relationships to GY and GPRO as for the genomic estimated breeding values for indices calculated in the univariate models. A set of selection strategies generate sufficient predictive ability in phenotypic or genomic selection to be effective tools for simultaneous selection for GY and GPRO