Impacts of Population Structure and Analytical Models in Genome-Wide Association Studies of Complex Traits in Forest Trees: A Case Study in Eucalyptus globulus

Abstract

The promise of association genetics to identify genes or genomic regions controlling complex traits has generated a flurry of interest. Such phenotype-genotype associations could be useful to accelerate tree breeding cycles, increase precision and selection intensity for late expressing, low heritability traits. However, the prospects of association genetics in highly heterozygous undomesticated forest trees can be severely impacted by the presence of cryptic population and pedigree structure. To investigate how to better account for this, we compared the GLM and five combinations of the Unified Mixed Model (UMM) on data of a low-density genome-wide association study for growth and wood property traits carried out in a Eucalyptus globulus population (n = 303) with 7,680 Diversity Array Technology (DArT) markers. Model comparisons were based on the degree of deviation from the uniform distribution and estimates of the mean square differences between the observed and expected p-values of all significant marker-trait associations detected. Our analysis revealed the presence of population and family structure. There was not a single best model for all traits. Striking differences in detection power and accuracy were observed among the different models especially when population structure was not accounted for. The UMM method was the best and produced superior results when compared to GLM for all traits. Following stringent correction for false discoveries, 18 marker-trait associations were detected, 16 for tree diameter growth and two for lignin monomer composition (S∶G ratio), a key wood property trait. The two DArT markers associated with S∶G ratio on chromosome 10, physically map within 1 Mbp of the ferulate 5-hydroxylase (F5H) gene, providing a putative independent validation of this marker-trait association. This study details the merit of collectively integrate population structure and relatedness in association analyses in undomesticated, highly heterozygous forest trees, and provides additional insights into the nature of complex quantitative traits in Eucalyptus.

Linkage and Association Mapping for Two Major Traits Used in the Maritime Pine Breeding Program: Height Growth and Stem Straightness

“…This consideration of relatedness considerably reduced p -value inflation, for both traits. Similar results were reported for wood quality and growth traits in Eucalyptus globulus [70, 71] and for wood property traits in Cryptomeria japonica [72]. Only seven of the 2,498 SNPs, from six different loci, were found to be significantly associated with the traits considered (p < 0.001).…”

Independent and Joint‐GWASfor growth traits inEucalyptusby assembling genome‐wide data for 3373 individuals across four breeding populations

“…Although less genetic variation is in principle available in such structured populations, the associations detected in genetically improved material should be considerably more relevant to further breeding, as their effect would be relevant in an already elite background. Following an equivalent rationale, three studies have reported GWAS for growth and wood quality traits in breeding populations of Eucalyptus (Cappa et al ., ; Müller et al ., ; Resende et al ., ). Interestingly, the results were on par with those described earlier in natural populations, in which few associations explaining small fractions of the genetic variation were detected.…”

A flexible multi‐species genome‐wide 60K SNP chip developed from pooled resequencing of 240 Eucalyptus tree genomes across 12 species

“…In recent years, however, hybridization‐based DArT arrays for eucalypts have supplied on average 3000 informative markers across species, increasing by an order of magnitude the ability to query polymorphisms across the genome (Sansaloni et al ., ). This has inaugurated genome‐wide mapping (Hudson et al ., ; Petroli et al ., ), QTL detection (Freeman et al ., ), association genetics (Cappa et al ., ) and genomic selection in eucalypts (Resende et al ., ). Nevertheless, the dominant behavior of DArT markers, together with limitations in expanding their number and distributing them equally across the genome, became concerns when considering large‐scale, fast turnaround genotyping for operational molecular breeding.…”