Genetic Variation and Genome-Enabled Prediction of White Lupin Frost Resistance in Different Reference Populations

In various European regions, white lupin (<i>Lupinus albus</i> L) production could increase by autumn sowing of winter-hardy varieties. This study aimed to explore the genetic variation, the genetic architecture, and the genomic prediction of frost resistance in two reference populations, one including 144 landrace and cultivar genotypes, and the other comprising 144 breeding lines. These populations were genotyped by 40,914 and 32,951 SNP markers, respectively, issued by genotyping-by-sequencing. The genotypes were phenotyped for mortality and a biomass injury score at freezing temperature of -11 °C under controlled conditions. Both traits, highly correlated, exhibited large genetic variation and high broad-sense heritability (<i>H</i>² = 0.76-0.82). A genome-wide association study highlighted their polygenic architecture and detected markers linked to candidate genes. The intra-population predictive ability of plant mortality achieved 0.41 for landrace and cultivar germplasm, and 0.67 for breeding lines. The cross-population predictive ability was higher when using the model constructed for landrace and cultivar germplasm to predict breeding lines (0.39) than the reverse (0.26). Landrace field survival was largely influenced by late phenology in addition to frost resistance. Our results revealed frost-resistant germplasm, confirmed the polygenic control of frost resistance, and highlighted genomic prediction opportunities for line selection and the identification of elite genetic resources.