Breeding forest trees by genomic selection: Current progress and theway forward

Abstract

A challenge common to all forest tree improvement programs is the long time interval of a breeding cycle. Moreover, the large size of trees, late trait expression and the extended time-lag between the breeding investment and the deployment of genetically improved material, make tree breeding a costly operation, more susceptible to changes in market demands, business objectives and climate change. The outlook of accelerating tree breeding and improving selection precision by marker assisted selection (MAS), thus became one of the driving principles of most forest tree genome projects. Although important advances were made in quantitative trait locus (QTL) mapping and association genetics, MAS did not make it in the ’real tree breeding world’. Limitations of early genomic technologies, coupled to the genetic heterogeneity of tree species and an overoptimistic assessment of the architecture of complex traits in such phenotypically plastic perennial organisms, largely explain this outcome. The inability to ascertain and make use of individual QTLs has caused a paradigm shift from trying to dissect trait components and determine their individual effects, to dealing with the aggregate of whole-genome effects to predict phenotypes by Genomic Selection (GS). Given the rapidly growing interest of tree breeders on this theme, this chapter provides an update on the current status and upcoming perspectives of GS in forest tree breeding. After a brief explanation of the basic principles and the main factors that impact prediction accuracy, the perspectives and the encouraging experimental results of GS in forest trees are reviewed. Concerns raised by tree breeders about GS are then discussed by reviewing the current knowledge in other species, while attempting to provide a roadmap for upcoming research and operational applications of GS. The prospects of GS in tree breeding are very promising to increase genetic gain per unit time through improved estimation of breeding (parent selection) and genotypic (clone selection) values, reduction of generation time and optimization of genome-directed mate allocation. Furthermore, the progressive accumulation of huge genotype and corresponding phenotype datasets in GS will provide an exceptional ’big data’ framework that should enhance our understanding of the connection between genome-wide elements and the observable phenotypic variation in complex traits.