Genomic Dissection and Prediction of Heading Date in Perennial Ryegrass

So far, the implementation of Genomic Prediction (GP) in plant breeding has been mainly investigated in crops farmed as genetically homogeneous cultivars. The aim of this work is to test the feasibility of GP in forage perennial ryegrass (Lolium perenne L.), using breeding material from different steps of the breeding program.

Heading date was used as a model trait, due to its ease of assessment and relatively high heritability. Genomic data was produced by Genotyping-By-Sequencing of family bulk samples, and mean allele frequencies were estimated per family at each marker. Phenotype data was also collected on family means from sward plots.

Analyses involved: (i) Genome Wide Association study, within F2 families, in order to uncover the genetic architecture of the trait; (ii) estimation of the breeding values in a set of related synthetics, performed both by Marker Assisted Selection (MAS) and through GP; (iii) comparison of different models for GP within F2 families, using different cross-validation schemes, designed to test specific hypothesis.

A total 19 markers were significantly associated with heading date, some of them locating within or proximal to genes involved in floral regulation (e.g. CONSTANS and Phytochrome C). Other markers were situated in the intergenic region, revealing a relatively complex genetic architecture. GP models yielded very accurate predictions, significantly superior to those obtained with MAS using the 19 significant markers. Higher accuracies were found when predicting from related families than when predicting families unrelated to families used to train the model.