Selecting multi-use naked barley for Irish organic systems

Organic farming has gained popularity among consumers, policymakers, and farmers due to its perceived health and environmental benefits. In organic cereal crop production, restrictions on chemical inputs may also lead to higher variability in grain yield and product quality across years and locations. Growing locally adapted crop species and varieties is the key to sustainable yield increase. Barley shows good performance in the Irish soil and climate conditions, and the naked type, thanks to the easy hull removal, shows more feed, food, and malt applications than the hulled type. This study was undertaken to develop locally adapted naked barley varieties with multi-use properties and assess the ability of new breeding techniques, such as genomics and phenomics, to accelerate crop improvement. A large panel of 247 spring naked barley accessions was evaluated under Irish organic conditions over three growing seasons (2019, 2020, 2021), following a type II modified augmented experimental design (MAD2), with data collected on traits related to phenology, agronomy, diseases, threshing ability, and grain quality. The accessions were sourced from various breeding programs and genebanks and exhibited diverse phenotypic properties such as spike morphology and seed colour. The Multi-trait Genotype-Ideotype Difference Index (MGIDI) allowed the selection of varieties with good performance and stability. Moreover, results suggested adding diseases, green leaf, and phenological traits to the ideotype for improved long-term performance stability on yield and grain quality. Eighteen Vegetation Indices (VIs) were calculated from RGB and multispectral reflectance data. The VIs' genotypic variation and correlation with ground-measured traits were only significant for recordings from booting to anthesis development stages. The vegetation coverage index (VCOV) was the most robust VI, showing the highest correlation with yield-related traits. However, the study highlighted poor consistency of the data across flights and thus the need for standardized measurement protocol and camera sensor. Genome-wide association Studies (GWAS) and Genomic Prediction (GP) were carried out based on the genotyping information on 50k Single Nucleotide Polymorphisms (SNPs) for each accession. Four GWAS models, namely Gmodel, BLINK, EMMA, and 3VMrMLM, were applied to the single-year, mean, and multi-year datasets. GWAS identified 2034 Marker-Trait Associations (MTAs), including 332 discovered in at least two analyses, of which 72 explained more than 5% of the random trait variance. Twenty reliable genes were identified for straw strength, plant height, grain plumpness, and threshability, mainly discovered with 3VMrMLM. 3VMrMLM also took good account of the epistasis effect, with the number of favourable alleles correlated with the phenotypic value. Therefore, genotype selection based on this parameter may favour long-term genetic gain than based on candidate genes or phenotypic data. The Bayesian Reproducing Kernel Hilbert Space (RKHS) GP model was applied to aggregated data from 2020 and 2021, testing a single-trait model with yearly data as multiple traits (ST) and a multi-trait model with year as a fixed factor (MT). Adding GWAS main-effect markers improved the predictive ability (PA) and model fit in all cases. Predicting tested lines with available phenotypic data on one year (CV2) showed higher PA than with no data (CV1). Correcting for population structure significantly increased PA for ST-CV2 but led to similar or lower PA otherwise. None of the strategies significantly improved yield predictive ability. Overall, the study demonstrated the potential of MAD2 design to effectively screen a large and diverse panel of naked barley lines under organic conditions. Additionally, phenomics and genomics can be valuable tools for more cost-efficient breeding, provided that adequate measurement protocols and modelling strategies are applied.