Genetic parameter estimates informed by a marker-based pedigree: A case study with Eucalyptus cladocalyx in southern Australia

Abstract

Analysis of stem diameter, height and axis persistence (AP) in a first-generation Eucalyptus cladocalyx breeding population comprising 137 wild and land-race families planted at 11 sites in southern Australia revealed significant genetic variation among subpopulations and among families within subpopulations. Alternative analyses were carried out using individual-tree mixed models that (i) assumed the trees within families were half-siblings (HS) and (ii) used molecular marker-based information to account for highly heterogeneous relatedness and inbreeding depression among families resulting from mixed mating (MM). For certain site and trait combinations, the HS models would not converge, as estimates of additive variance exceeded the total phenotypic variance, demonstrating the fundamental unsuitability of the HS assumption for this breeding population. Where HS models converged, moderate to very high heritability estimates resulted for growth traits. The MM assumption resulted in re-ranking of individual-tree breeding values and markedly lower estimates of narrow-sense heritability for all trait-site combinations. In some cases, however, heritability remained moderate to high, probably reflecting unquantified dominance variance in some highly inbred subpopulations. Genotype-by-environment interaction was significant overall due to reactivity of genotypes on a few sites, with type-B correlations between pairs of sites ranging from 0.06 to 0.99. Generally, families from the Australian land race were found to perform particularly well for both growth and AP traits. Some wild families were found to be vigorous, despite significant inbreeding. The study has demonstrated that traditional models assuming non-relatedness and/or homogeneous inbreeding in first-generation eucalypt breeding populations can be significantly improved upon by flexible mixed models that integrate marker-based data. Show more