Contribution of white clover hosts to the effectiveness of rhizobia symbioses : A thesis submitted in partial fulfilment of the requirements for the Degree of Doctor of Philosophy at Lincoln University

White clover (Trifolium repens L.) is an integral component of mixed pastures in New Zealand temperate agriculture, providing quality feed and a sustainable source of fixed nitrogen (N) through its symbiosis with strains of Rhizobium leguminosarum bv. trifolii. While there has been much focus on identifying and applying Rhizobium strain inoculants for improved N-fixation, there has been less attention on identifying and exploiting plant genetic factors to develop cultivars that can consistently form effective Rhizobium symbioses. The complexity of the clover-Rhizobium interaction makes for a challenging but valuable breeding target which represents an important strategy for increasing the sustainability of New Zealand’s pasture-based agriculture. In this study, the phenotypic variation of N-fixation capacity across individuals from 17 white clover cultivars, including a wild ecotype (‘Tienshan’) and a T. repens × T. uniflorum interspecific hybrid cultivar (‘AberLasting’), released between 1920 and 2003 and representing three leaf size classes were investigated. N-fixation capacity was assessed by growing plants in vermiculite/McKnight’s solution, where N was provided either by an effective commercial strain (TA1) or partially effective (S12N10) N-fixing Rhizobium strain or supplied mineral N (positive control). Symbiotic traits recorded after 35 days of growth included Shoot and Root dry matter (DM), root to shoot ratio (RSR), and symbiotic potential (SP) which is the proportion of biomass generated through symbiosis relative to plants grown with mineral N. There were significant differences or interactions between and among cultivars and Rhizobium strains. Furthermore, there was a high degree of variability in N-fixation capacities of individuals within the cultivar populations. Multivariate analysis clustered cultivars across an N-availability gradient and identified some cultivars that performed equally well (‘S100’) or equally poorly (‘Barblanca’) or had marked differential performance (‘Dutch’ and ‘Tienshan’) across effective or partially effective strains. In order to derive key quantitative genetic parameters which could be utilised for improving N-fixation in white clover, a genetically structured F3 population of half-sibling (HS) families (n = 120) underwent the same screening assay as in the previous study, with N being provided either by the commerical strain TA1 or supplied mineral N (positive control). Shoot and Root DM, SP as well as RSR were measured and differed considerably among HS families, with 44% of families forming sub-optimal symbioses with TA1. Best linear unbiased predictor (BLUP) means for all traits were derived for each HS family and used to estimate significant additive variation and narrow-sense heritabilities, which ranged from 0.05 (RSR) to 0.33 (Shoot DM). This indicated there is additive genetic variability that can be exploited for improving the symbiotic interactions. These estimates also enabled the genetic gain for the examined traits to be predicted at 20%, 10% and 5% selection pressures while also identifying 24 elite HS families that could be utilised in future selection validation work. The feasibility of applying genomic selection (GS) for improving genetic gain in white clover-Rhizobium associations using a HS family structure was investigated. The phenotypes of 118 HS families and a genomic relationship matrix derived from ~155,000 genotyping-by-sequencing single nucleotide polymorphism markers of the HS families’ corresponding maternal parents were used for genomic prediction (GP) using five models. Cross-validation predictive ability (PA) of symbiotic traits ranged from 0.14 (Root SP) to 0.36 when two traits (Shoot DM and Shoot SP) were combined into a multi-trait Smith-Hazel selection index. The PAs of the five GP models were similar across all traits and therefore the computationally efficient GP model Kinship using GBS with depth adjustment-GBLUP (KGD-GBLUP) was selected. Different parameters influencing PA were assessed and identified a minimum of 94 HS families using a training: test combination of 80: 20%, respectively, over 100 iterations was sufficient for generating PAs in this population. This work provides an ideal foundation to assess genomic selection for breeding white clover germplasm for improved N-fixation and productivity by undertaking among-(using phenotype) and-within (using GS) HS family selection. To determine relationships between performance with pure strains and soil communities, a subset of cultivars (n = 4) and HS families (n = 4) screened previously for differential N-fixation capacities were assessed with two pure strains (TA1 and S12N10) and five soil rhizobia extracts. Differences in the N-fixation capacities with pure strains reflected those from the previous experiments. While there was considerable variation in symbiotic traits among populations with the five soil extracts, performance with the effective Rhizobium strain TA1 was often indicative of performance with the soil communities. Using multivariate pattern analysis, distinct clusters among the white clover populations were identified, which appeared to be driven by the number and appearance of nodules. In summary, the findings of this thesis will provide valuable information for facilitating the improvement of N-fixation in white clover in future breeding efforts.