Epistasis in wheat breeding

Abstract

The core objective of wheat breeding is to develop superior varieties to the target population of environments. As most agronomic traits are quantitative, the underlying genetic architecture is assumed to be additive, implying that the superiority of parental lines is transmitted to the offspring. However, often rather than rarely, wide crosses between elite genotypes yield poor breeding populations, indicating that additive might not be the major type of gene action. The inbreeding nature of wheat suggests that epistasis could be the underlying genetic architecture. The work presented here describes conventional and innovative methods to test the existence and extent of epistasis in wheat breeding. The first chapter presents a comprehensive literature review that discusses the theoretical framework for the creation of epistasis from gene duplication events and the expansion of this construct to the population levels. Then, the history of wheat is utilized to discuss the evolution of epistasis into a complex genetic system and the strategies utilized by breeding programs to manage this complexity and attain satisfactory levels of genetic gain. The second chapter utilizes an innovative approach to test the existence of sign epistasis in wheat breeding. Sign epistasis occurs when the effect of an allele can be beneficial or detrimental upon variation in the genetic background. Through the coupling of interchromosomal linkage disequilibrium analysis with the calculation of allele frequencies, 19 candidate interactions of sign epistasis were identified. Although the validation analyses attributed to random genetic drift the sign epistasis patterns observed in 11 candidate interactions, eight interactions presented strong evidence for sign epistasis and the main hypothesis could not be refuted. These findings imply that the effect of alleles can vary from positive to negative depending on the genetic background and explain the often poor combining ability of high performing lines. The third chapter explores the genetic architecture of wheat grain yield by modeling additive and additive-by-additive epistatic effects, using data from 3740 experimental lines. Modeling epistasis in whole genome or sub genome models marginally improved prediction accuracy in genomic selection models and resulted in non-orthogonal partition of genetic variance components. The estimation of sub genome additive effects showed that the best lines did not have the greatest additive effects in more than one sub genome, suggesting the possibility of making targeted crosses that combine genotypes with the highest additive effects in each subgenome.