Dissection of the genetic architecture of three seed-quality traits and consequences for breeding in Brassica napus

Abstract

Genome-wide association studies (GWASs) combining high-throughput genome resequencing and phenotyping can accelerate the dissection of genetic architecture and identification of genes for plant complex traits. In this study, we developed a rapeseed genomic variation map consisting of 4 542 011 SNPs and 628 666 INDELs. GWAS was performed for three seed-quality traits, including erucic acid content (EAC), glucosinolate content (GSC) and seed oil content (SOC) using 3.82 million polymorphisms in an association panel. Six, 49 and 17 loci were detected to be associated with EAC, GSC and SOC in multiple environments, respectively. The mean total contribution of these loci in each environment was 94.1% for EAC and 87.9% for GSC, notably higher than that for SOC (40.1%). A high correlation was observed between phenotypic variance and number of favourable alleles for associated loci, which will contribute to breeding improvement by pyramiding these loci. Furthermore, candidate genes were detected underlying associated loci, based on functional polymorphisms in gene regions where sequence variation was found to correlate with phenotypic variation. Our approach was validated by detection of well-characterized FAE1 genes at each of two major loci for EAC on chromosomes A8 and C3, along with MYB28 genes at each of three major loci for GSC on chromosomes A9, C2 and C9. Four novel candidate genes were detected by correlation between GSC and SOC and observed sequence variation, respectively. This study provides insights into the genetic architecture of three seed-quality traits, which would be useful for genetic improvement of B. napus.

Keywords: candidate genes; genome-wide association study (GWAS); genomic variation; quantitative trait loci (QTLs); rapeseed (Brassica napus); seed-quality traits.

Figure 1

Heatmap for distribution of genomic variations in rapeseed. The number of SNPs and INDELs was calculated by 100‐kb sliding window across each chromosome.

Figure 2

Genome‐wide association studies of three seed‐quality traits. Manhattan plots for EAC (a), GSC (b) and SOC (c) across multiple environments. ‐log₁₀(p) values are plotted against position on each chromosome. Dashed grey lines indicate the genome‐wide significance threshold.

Figure 3

Characterization of PVE and favourable allele of associated loci for three traits. (a) The frequencies of favourable alleles and corresponding PVE. The favourable alleles of each loci for EAC and GSC were that reducing EAC and GSC in seeds, respectively. The favourable alleles of each loci for SOC were that increasing SOC. (b) Total PVE of all significant loci for three traits across each environment. Numbers on top of each bars indicated the number of loci detected in each year. (c‐e) Correlation between favourable allele number and EAC, GSC and SOC, respectively.

Figure 4

Associated loci and candidate genes for EAC on chromosome A8 and C3. (a, d) Regional Manhattan plot surrounding the peak signals on chromosome A8 (a) and C3 (d) in 2011. Red dot indicates the peak signals snp1348001 (a) and snp2429000 (d). Dashed line represents the significance threshold. (b, e) Exon structure and functional variations of BnaA08g11130D (b) and BnaC03g65980D (e). Numbers indicate the positions of open reading frame from translation start site. (c, f) Boxplots for EAC based on the genotypes of BnaA08g11130D (c) and haplotypes of BnaC03g65980D (f). Differences between the genotypes or haplotypes were analysed by Wilcoxon rank‐sum test. n.s. represents not significant.

Figure 5

Associated loci and candidate genes for GSC on chromosome C9 and A9. (a, d) Regional Manhattan plot surrounding the peak signals on chromosome C9 (a) and A9 (d) in 2011. Red dot indicates the peak signal indel479620 (a) and indel240140 (d). Dashed line represents the significance threshold. (b, e) Exon–intron structure and functional variation of BnaC09g05300D (b) and BnaA09g08410D (e). Numbers indicate the positions of open reading frame from translation start site. (c, f) Boxplots for GSC based on the genotypes of BnaC09g05300D (c) and BnaA09g08410D (f). Differences between the genotypes were analysed by Wilcoxon rank‐sum test.

Figure 6

Associated loci and candidate genes for SOC on chromosome A10 and C4. (a, d) Regional Manhattan plot surrounding the peak signals on chromosome A10 in 2010 (a) and C4 in 2012 (d). Red dot indicates the peak signals snp1778098 (a) and snp2649172 (d). Dashed line represents the significance threshold. (b, e) Exon–intron structure and functional variations of BnaA10g23290D (b) and BnaC04g45690D (e). Numbers indicate the positions of open reading frame from translation start site. (c, f) Boxplots for SOC based on the haplotypes of BnaA10g23290D (c) and BnaC04g45690D (f). Differences between the genotypes or haplotypes were analysed by Wilcoxon rank‐sum test. n.s. represents not significant.