GBS-Based Yield Selection in Alfalfa: Now We're Getting Somewhere...Aren't We?

Alfalfa (Medicago sativa), a major forage crop worldwide, is perennial, outcrossing, and autotetraploid. Breeding cycles typically take two to five years to complete, limiting genetic gain compared to annual crops. Consequently, the potential for genomic selection to accelerate yield improvement appears high. We have developed two high throughput genotyping platforms for alfalfa – an Infinium array of 10,000 features and genotyping-by-sequencing – that effectively cover the alfalfa genome and make genomic selection feasible, at least conceptually. We have developed genetic linkage maps in diploid and tetraploid alfalfa populations and shown that the GBS markers are largely collinear with the whole genome sequence of the related species, M. truncatula. Because GBS marker data is cheaper to generate than array data, we are using GBS for genomic selection. We have applied GBS to alfalfa breeding populations and have been able to estimate prediction accuracies within locations and also between cycles of selection. In a clonal selection experiment, accuracies of 0.30-0.50 were found for yield of individual harvests within locations and across harvests and locations; intergenerational prediction accuracies were as high as 0.40. Accuracies for breeding values of parents evaluated from half-sib progeny yield responses were around 0.25 in one data set and 0.35 in another. These results suggest that genomic selection could offer potential for improving yield. We will discuss the exciting opportunities that these results provide, and offer suggestions on how to best apply this technology to alfalfa breeding programs.