Identifying optimum breeding methods for improving seasonal yields and vegetative persistence in red clover : A thesis submitted in partial fulfilment of the requirements for the Degree of Doctor of Philosophy at Lincoln University

Abstract

Worldwide, red clover (Trifolium pratense L.) is an important pastoral legume species that can produce forage high in protein, is readily digestible and can grow in a range of soil types and environmental conditions. Red clover can fix atmospheric nitrogen biologically to generate plant-available nitrogen through symbiosis with soil-dwelling Rhizobium bacteria due to being a leguminous species. Red clover is grown either as a pure stand for a cut and carry farming system or in mixed sward pastures. Primarily grown in temperate regions as a mixed sward, white clover (Trifolium repens L.) and perennial ryegrass (Lolium perenne L.) are typically the most common companion species for red clover. The origin of red clover is thought to be the Eastern Mediterranean region, but it is now present throughout most temperate regions globally with this genetic diversity captured in germplasm collections worldwide. The backbone of this thesis was a panel of 92 geographically diverse germplasm populations sourced from the Margot Forde Genebank (AgResearch Grasslands Research Centre, Palmerston North, New Zealand). When selecting these populations, particular consideration was given to populations from regions of lower rainfall. Overall, the selected populations represented 15 countries. The genetic diversity, population structure, agronomic performance and heritability of key morphological and physiological traits of these populations was assessed in a glasshouse pot trial, multi-year multi-site field trials,and genotyping-by-sequencing (GBS). The development of linear mixed models on genomic, phenotypic, and environmental information meant variance components and genotype-by-environment (G x E) interactions for key morphological traits for the 92 populations were assessed. Key relationships between plant performance and environmental variable were also evaluated. Discriminant analysis of principal components (DAPC) assigned the 92 populations into seven clusters, with a strong alignment shown to geographic origin. Country of origin not only influenced the genetic structure of the populations, but also the expected heterozygosity, a measure for genetic diversity within populations, which ranged from 0.08 to 0.17. High narrow-sense heritability (h2 > 0.70) was found for eight morphological traits and the influence of mean precipitation, temperature and isothermality (variance in daily temperature relative to annual variation) of the original collection locations, on plant and trait performance was also highlighted. This also included identification of genes associated with these bioclimatic variables, which have the potential to serve as genetic markers for future use. Further analysis was applied to better identify and understand bioclimatic variables driving environmental adaptation and the resulting DNA variants (outlier single nucleotide polymorphisms (SNPs)) associated with adaptation. The calculation of adaptive indices and genomic offset values allowed the identification of the agronomic suitability of the 92 populations for future NZ environments. Annual mean diurnal range, isothermality, mean temperature of the wettest quarter, and precipitation seasonality were found to influence adaptive genetic variation most. Forty-two outlier SNPs strongly associated with key bioclimatic variables showed potential as markers for climate-resilient breeding. Predictive mapping and genomic offset values highlighted the suitability of the 92 populations to provide important genetic diversity to develop future adapted cultivars. Seven of the 92 populations were tested for performance under water deficit and waterlogged conditions in a controlled glasshouse experiment. Twelve traits were used to evaluate plant responses to water stress. Under water deficit, all aboveground morphological traits were affected, highlighted by decreases in total dry matter (40%), leaf number (50%) and leaf thickness (50%). Additionally, an increase in the root to shoot ratio indicated a shift to maintaining root biomass, a trait attributed to plant water deficit tolerance. Under waterlogged conditions, a reduction in photosynthetic activity was accompanied by decreases in plant performance, root dry mass (83%), total dry matter (30%) and leaf number (34%). The need for trait improvement under both water stresses was underlined by this experiment. Of the 92 populations evaluated in multi-year, multi-site field trials, 12 populations along with two cultivar controls were further evaluated. Plant yield, plant persistence and root structure were assessed, using thirteen above- and below-ground traits. The biomass production of the germplasm populations was significantly lower than the cultivar controls. Root structure between the populations varied with germplasm populations, having either an expansive or compact root system, whereas the controls were found to have a mixture of both. Additional key relationships between root structure and both plant persistence and plant production were identified. This highlighted the variation in trait expression above- and below-ground and highlighted the benefits of finding the right balance of traits for production or survival. After assessing genetic diversity among the 92 populations, the potential to introgress novel germplasm into locally adapted populations was evaluated by forming two cultivar-derived sets of half-sib families, each two generations (F2) removed from the original material. Plant performance and persistence was tested in multi-location, multi-year field trials. By integrating genomic, phenotypic, and environmental data, we identified significant SNPs and evaluated their roles in key trait expression, while also developing genomic prediction models to estimate prediction accuracy across traits. Several half-sib families performed as well as or better than the locally adapted populations across sites. In total, 35 SNPs and 27 associated genes were linked to phenotypic trait expression, with growth and plot density traits showing particularly high prediction accuracy. The evaluation of introgressed material, discovery of molecular markers, and development of genomic prediction models highlighted the potential to efficiently identify and incorporate individuals carrying desirable traits into breeding programs. Taken together, these studies provide valuable insight into the genetic diversity present in germplasm populations, how key traits are inherited and how they perform in new environments, along with how environmental variables interact with trait expression and plant adaptability. The key findings from this thesis will contribute to improved breeding programs for developing future adaptive cultivars through the utilization of novel germplasm populations.