Hybridisation-based target enrichment of phenology genes to dissect the genetic basis of yield and adaptation in barley

Abstract

Barley (Hordeum vulgare L.) is a major cereal grain widely used for livestock feed, brewing malts and human food. Grain yield is the most important breeding target for genetic improvement and largely depends on optimal timing of flowering. Little is known about the allelic diversity of genes that underlie flowering time in domesticated barley, the genetic changes that have occurred during breeding, and their impact on yield and adaptation. Here, we report a comprehensive genomic assessment of a worldwide collection of 895 barley accessions based on the targeted resequencing of phenology genes. A versatile target-capture method was used to detect genome-wide polymorphisms in a panel of 174 flowering time-related genes, chosen based on prior knowledge from barley, rice and Arabidopsis thaliana. Association studies identified novel polymorphisms that accounted for observed phenotypic variation in phenology and grain yield, and explained improvements in adaptation as a result of historical breeding of Australian barley cultivars. We found that 50% of genetic variants associated with grain yield, and 67% of the plant height variation was also associated with phenology. The precise identification of favourable alleles provides a genomic basis to improve barley yield traits and to enhance adaptation for specific production areas.

Keywords: Hordeum vulgare; association mapping; flowering time; grain yield; next-generation sequencing; target capture.

Figure 1

Plot of ancestry estimates inferred by ADMIXTURE for 895 worldwide barley accessions for 4260 SNPs. (a) Each colour represents a population, and the colour of individual haplotypes represents their proportional membership in the different populations. Distribution of ADMIXTURE‐defined populations based on (b) three growth habits, (c) two row types, and (d) seven geographical locations. FAC, facultative; SPR, spring; WIN, winter growth habit.

Figure 2

PhyC genomic region shows strong association signals for phenology and grain yield. (a) Manhattan plot for chromosome 5H with association signals for phenology (days to Z49, n = 462) highlighted in green. GWAS results are presented by negative log₁₀ of FDR adjusted P‐values (q‐values) against position on the chromosomes (n = 462). Horizontal dashed lines indicate the genome‐wide significant threshold selected by local false discovery rate and a q‐value cut‐off at 0.05 (blue) and 0.01 (red). (b) Summary of local LD and haplotype blocks for the PhyC genomic region containing all 21 detected SNPs. LD plot, generated in Haploview, indicates r ² values between pairs of SNPs multiplied by 100; white, r ² = 0; shades of grey, 0 < r ² < 1; black, r ²  = 1. Haplotype blocks (blocks 1–3) in the PhyC genomic region were defined with the four‐gamete rule method. The twelve SNPs that were highly significantly associated with days to Z49 in the optimal MLM are highlighted in red font. (c) The diagrammatic structure of the conserved domains on exon 1 of PHYC and location of two variants detected within this region. (d) Days to Z49 variation between different genotypes for Chr_5_598560301_T/C, and Chr_5_598561262_T/C. (e) Grain yield variation between different genotypes for Chr_5_598560301_T/C, and Chr_5_598561262_T/C. P‐values calculated using Kruskal‐Wallis tests. ****P value <0.0001.

Figure 3

AGLG1 genomic region shows strong association signals for phenology, grain yield, and plant height. (a) Manhattan plot for chromosome 5H with association signals for flowering time (days to Z49) highlighted in green. (b) Manhattan plot for chromosome 5H with association signals for grain yield (kg/ha) highlighted in green. (c) Manhattan plot for chromosome 5H with association signals for plant height (cm) highlighted in green. All GWAS results are presented by negative log₁₀ of FDR adjusted P‐values (q‐values) against position on each of the seven chromosomes (n = 462). Horizontal dashed lines indicate the genome‐wide significant threshold selected by local false discovery rate and a q‐value cut‐off at 0.05 (blue) and 0.01 (red). (d) Summary of local LD and haplotype blocks for the AGLG1 genomic region containing all 21 detected SNPs. LD plot, generated in Haploview, indicates r ² values between pairs of SNPs multiplied by 100; white, r ² = 0; shades of grey, 0 < r ² < 1; black, r ² = 1. Haplotype blocks (blocks 1–3) in the AGLG1 genomic region were defined with the four‐gamete rule method. SNPs that were highly significantly associated with phenology in the optimal MLM are highlighted in red font. (e) Days to Z49 variation between different genotypes for Chr_5_599329482_A/T, and Chr_5_599333006_C/T. (f) Grain yield variation between different genotypes for Chr_5_599329482_A/T, and Chr_5_599333006_C/T. (g) Plant height variation between different genotypes for Chr_5_599329482_A/T, and Chr_5_599333006_C/T. P‐values calculated using Kruskal‐Wallis tests. ****P value <0.0001.