Genotyping-by-Sequencing and Its Exploitation for Forage and Cool-Season Grain Legume Breeding

Genotyping-by-Sequencing (GBS) may drastically reduce genotyping costs compared with single nucleotide polymorphism (SNP) array platforms. However, it may require optimization for specific crops to maximize the number of available markers. Exploiting GBS-generated markers may require optimization, too (e.g., to cope with missing data). This study aimed (i) to compare elements of GBS protocols on legume species that differ for genome size, ploidy, and breeding system, and (ii) to show successful applications and challenges of GBS data on legume species. Preliminary work on alfalfa and Medicago truncatula suggested the greater interest of ApeKI over PstI:MspI DNA digestion. We compared KAPA and NEB Taq polymerases in combination with primer extensions that were progressively more selective on restriction sites, and found greater number of polymorphic SNP loci in pea, white lupin and diploid alfalfa when adopting KAPA with a non-selective primer. This protocol displayed a slight advantage also for tetraploid alfalfa (where SNP calling requires higher read depth). KAPA offered the further advantage of more uniform amplification than NEB over fragment sizes and GC contents. The number of GBS-generated polymorphic markers exceeded 6,500 in two tetraploid alfalfa reference populations and a world collection of lupin genotypes, and 2,000 in different sets of pea or lupin recombinant inbred lines. The predictive ability of GBS-based genomic selection was influenced by the genotype missing data threshold and imputation, as well as by the genomic selection model, with the best model depending on traits and data sets. We devised a simple method for comparing phenotypic vs. genomic selection in terms of predicted yield gain per year for same evaluation costs, whose application to preliminary data for alfalfa and pea in a hypothetical selection scenario for each crop indicated a distinct advantage of genomic selection.