Chapter 8 General conclusions
8.1 Results and their application to plant breeders
8.1.2 Adoption of novel breeding methods
The objective of plant breeding is to genetically improve the performance of a species, which in this situation is white clover, in the most efficient manner possible (Fehr, 1987).
Development of an efficient strategy hinges on the selection of an appropriate breeding method coupled with the thoughtful allocation of resources for population development and genotype selection (Fehr, 1987). Adoption of the most efficient method therefore requires prior information on the amount of genetic improvement a range of alternative methods can achieve, within a given resource allocation.
The ideal scenario to compare alternative breeding methods would be to empirically measure realised genetic gain in terms of the mean performance of a population among a range of different methods. Unfortunately the empirical comparisons of different breeding methods is time consuming, laborious and in most cases far beyond practicality for most species. An obvious exception is maize, where alternative breeding methods have been compared empirically (Weyhrich et al., 1998). A common way for breeders to access the merits of alternative breeding methods is to compute the amount of genetic gain using prediction equations (Fehr, 1987). Genetic gain is often presented on a per year basis to account for the
177 variation in breeding method duration. Computation using prediction equations is based on genetic parameters estimated from genetic families representing random mating breeding populations. Similarly, the data generated from quantitative genetic experiments described in this thesis were able to be used to interpret the rationale for pursing various breeding methods in white clover.
In many situations, data from the experiments conducted in this thesis may help breeders to optimise or compliment current breeding strategies as opposed to replacing them. In terms of complementation to current breeding strategies, one way for breeders to employ findings from this thesis, is to include paternity testing into breeding methods as discussed below.
While the incorporation of paternity testing into breeding programmes such as lucerne (Riday et al., 2013) and red clover (Riday, 2011) requires little adjustment in breeding strategy since spaced-planted nurseries are already the norm (Riday, 2011), the progressive movement in white clover selection from mono-culture spaced-plants to duo-culture mini-plots (Woodfield and Caradus, 1994) makes the transition to paternity testing methods more problematic. For this reason, and the difficulty of maintaining individuality as discussed in Chapter 5, the use of paternity testing in white clover should be modified accordingly, to effectively transition into field breeding programmes. One such approach would be to use paternity testing to complement current plot systems as opposed to spaced-planted nurseries. While data can therefore not be collected on individual genotypes to determine paternal half-sib variances per se, since the best maternal families are also the best paternal families as shown in Chapter 5, there is no need to do so (assuming all paternal families are evaluated as maternal families).
Consequently, best maternal families would be identified, and then paternity testing would be used to help identify plants within plots that also have superior paternal genetics. The
combination of maternal and paternal selection in plots is similar to that demonstrated in spaced-planted nurseries, although selection within superior full-sib combinations based on phenotypic data is not available. No advantages are gained from paternal selection alone compared to maternal selection alone, since maternal and paternal genetic additive variances are similar (Chapter 5); contrary to the findings of Riday in red clover (2011).
Marker assisted selection (MAS) for within plot selection is highly beneficial in white clover, as a common problem associated with perennial forage species is the lack of plant identity within plots, making within-family selection relatively difficult. In many cases, breeders mitigate this difficultly of within plot selection by exposing trials to a greater range of abiotic and biotic factors to promote reductions in plant density. However this technique is only
178 useful if survivorship within families is low and heritability of vegetative persistence on a single plant basis is relatively high. At least paternity testing within plots or even remnant seed improves selection in this scenario by doubling the parental control factor based on replicated family data. The addition of low survivorship is also beneficial in these
circumstances, as less potential genotypes need to be screened with molecular markers for paternal identity. Traits with high heritability on an individual plant basis, such as leaf size and disease traits, may also be used as a preliminary criterion to put plants forward for paternity screening if plant density remains high.
Purely from a quality control perspective, DNA testing within plots enables the removal of any containment white clover volunteers as well, which despite the best practices of breeders, still act as contaminants via the high buried seed counts of most soils in New Zealand
(Clifford et al., 1990) and animal faeces (Suckling, 1950; Suckling, 1952).
As opposed to complementation of current breeding strategies, a second aspect of this thesis was to investigate alternative breeding strategies per se. This approach requires estimation of genetic parameters and simulation using prediction gain models, taking into account the magnitude and type of genetic variation (additive and non-additive variation). To date, the lack of literature available on the genetic parameters of random mating white clover populations evaluated in multiple target environments has limited the ability to simulate prediction gain models with realistic data (Jahufer et al., 2002). The heritabilities estimated in the experiments of this thesis provide breeders with improved understanding of heritabilities in actual farming systems and consequently allow breeders to critically evaluate their
breeding methodology more realistically, if these target environments are used to undertake selection.
Data generated from both the spaced-planted nurseries (Chapter 5) and the mini-plot trials (Chapter 6) clearly demonstrate the superiority of family selection methods for low heritable traits. By in large these low heritable traits include clover herbage yield, vegetative
persistence, number of nodes, number of rooted nodes and stolon branching. Traits where phenotypic recurrent selection methods are likely to be equally effective as or better than family selection methods include; leaf size, stolon thickness and growing point density. The benefit of family selection methods are well documented for low heritable traits as their data are obtained from replicated trials and therefore less affected by large environmental
variances (Nguyen and Sleper, 1983), which is demonstrated by the differentials in
179 heritabilities estimated on an individual plant basis and those on a family means basis in Chapter 5 and 6.
Among the family selection methods simulated, among-and-within-half-sib family selection (AWF-HS) according to the data generated in Chapter 5 (Figure 5.5) seems to be most the most logical choice for plant breeders to utilise if molecular markers are not available. These data are also supported by findings presented by Casler & Brummer (2008). AWF-HS selection allows breeders to accurately estimate the breeding worth of half-sib families for low heritable traits, as well as allowing for additional within family selection. The within family selection component would ideally suit traits of high heritability such as leaf size, growing point density, stolon thickness and simply inherited traits such as disease. The combination of replicated family selection for yield and vegetative persistence and within family selection (individual plant selection) for disease, leaf size and growing point density would fit commercial breeding programmes objectives well. The additional integration of molecular markers for paternity testing, such as among-half-sib-and-within molecular determined full-sib family selection (AWF-HS+MFS), would further elevate genetic gain above traditional AWF-HS, and allow more accurate within family selection for low heritable traits (maternal and paternal selection) such as herbage yield, in addition to the high heritable traits as well.
Crude estimates of additive and non-additive variances for most white clover attributes tend to agree with breeding methods which largely focus on additive genetic variation in white
clover, and forage species in general (Breese and Hayward, 1972). The possible exception to this could be herbage yield as shown in Chapter 7. This finding, while speculative due to the high associated standard error certainly warrants some further attention. Considering this, it would be worthy to note that progeny evaluation using full-sib testing may not be advisable, due to the biases of non-additive variation between families. If non-additive variation is indeed shown to significantly influence herbage yield in white clover, half-sib selection is a better option if the breeder is not interested in capturing the non-additive variation in hybrid cultivars.
As previously mentioned, when adopting half-sib evaluation, breeders must ensure random mating occurs within isolation cages. In addition, Vogel and Pederson (1993) highlighted the importance of conducting HSPT when the population is at linkage equilibrium. It was their belief that many breeders have sabotaged their own breeding success because parent
genotypes in their polycross nurseries came from various germplasm sources and were not in
180 linkage equilibrium, and it was highly probable that the half-sib progeny differences between families where due to differing levels of heterosis.
If the breeders opt to capture a proportion of the non-additive variation, the move towards hybrid evaluation and hybrid cultivars that are available for out-crossing forage species (Barrett et al., 2010; Michaelson-Yeates et al., 1997; Riday and Krohn, 2010) may be beneficial.