Genome-wide prediction of agronomic traits in hybrid spring-type canola (Brassica napus) using single nucleotide polymorphic (SNP) markers

Abstract

Canola/rapeseed (Brassica napus L., (AACC, 2n=38) is one of the world s most important oilseed crops and is used as human food, i.e. cooking oil and as animal feed. In Europe, winter-type canola is also used as a sustainable source of bioenergy. Canola was naturally formed ~7500 years ago from spontaneous inter-specific hybridisations between cabbage (Brassica oleracea) and turnip rape (Brassica rapa). Recently, the reference genome of the B. napus Darmor-bzh cultivar was sequenced and published in Science (Chalhoub et al. 2014) which provides new insights to be explored, to further improve this important oil crop in the coming time.Commonly used breeding materials of canola have developed a narrow gene pool due to continuous strong conscious selection by breeders for quality traits, i.e. low seed glucosinolate, low erucic acid, etc. Attempts have been made over the years to boost up the genetic diversity of canola through introgression from its progenitor species or other exotic materials. Breeders practice hybrid breeding in canola to exploit heterosis for improved agronomic traits, especially for seed yield gain and yield stability. Canola is considered to have a well-defined pollination control system, for example, cytoplasmic male sterility system (CMS), genic male sterility system (GMS), etc. and can be used for the production of hybrid seed with optimum success.Due to the recent advances in high-throughput genomic technologies, an avalanche of inexpensive single nucleotide polymorphism (SNP) markers is now available. These genome-wide markers have made molecular predictive breeding possible and lucrative in different crop species, i.e. Maize, rice, etc. I used the 60k Brassica SNP Illumina genotyping array in my study. After rigorous quality checks, a panel of single position 24,442 polymorphic SNPs distributed across the whole genome were used in my genomic analyses. First, I investigated the population structure in my dataset using the whole set of filtered SNP markers. Based on the K means clustering method, two main clusters along with one small cluster were diagnosed. I also explored chromosome-wise linkage disequilibrium (LD) decay within both the subgenomes A and C. The general pattern of LD was more conserved on C- subgenome than A- subgenome. This was in congruence with the previous reported studies in canola. Genome-wide association studies (GWAS) have emerged as a useful approach in genetics and has been used to correlate molecular markers with phenotypic variations in various crop populations. I used the GWAS approach to unravel genomic regions contributing to hybrid performance in canola and have identified candidate genes that have pleiotropic effects for two or more different traits. It has been reported already (Qian et al. 2007) that in canola hybrid breeding, additive gene effects are the main contributors to heterosis. General combining ability (GCA) accounts for additive gene effects. Therefore, GCA values were estimated for each pollinator in a set of 475 male lines and used in my genomic analyses instead of the per se F1 phenotype data. I used a mixed effects model approach which effectively accounts for the cryptic population structure. For GWAS, we considered three important agronomic traits, i.e. GCA for seed yield, GCA for DTF (days to flowering) and GCA for seed oil content.On chromosome A3, I found some Arabidopsis orthologue candidate genes with pleiotropic effects associated with significant SNP loci related to GCA for seed yield and GCA for DTF. For example, FRIGIDA (FRI) and EMBRYONIC FLOWER 2 genes which have been shown already in their direct role in flowering time and indirect role in yield related traits. Similarly, I reported a very important A.thaliana orthologue candidate gene (ABERRANT LATERAL ROOT FORMATION 1: ALF1) significantly associated with an overlapping SNP between GCA for seed yield and GCA for seed oil content. This gene is involved in various biochemical pathways, for example, unsaturated fatty acid biosynthetic process, adventitious root development, defense response, sulfur compound biosynthetic process and glucosinolate biosynthetic process. I also identified significant SNP haplotype diversity groups or blocks in the flanking regions of the significant SNPs in each trait that might contribute to trait heterosis in each of my examples. At the end, I reconstructed predicted F1 genotypes from the genotypes of the significant haplotypes from male lines (pollinators) and their corresponding haplotypes on the two tester lines (M1 and M2) in each trait. The genomic regions, candidate genes and the predicted F1 hybrid genotypes identified in my study provide worthwhile information that could be used in the future hybrid breeding strategies. My second project focused on the whole-genome prediction of hybrid performance in canola instead of identifying individual genes. Genome-wide selection (GS) or genomic prediction of unphenotyped germplasms (Meuwissen et al. 2001) is now rapidly making its way into plant breeding. In GS, molecular markers are employed across the whole genome simultaneously and genomic breeding values (GEBVs) are estimated. Pre-selection of the unphenotyped material is made on the basis of these GEBVs. Genomic prediction of test-cross hybrid performance in canola using widely-tested ridge-regression best linear unbiased prediction (RR-BLUP) model was carried out in this study taking seven important agronomic traits under consideration. These were seed yield (dt/ha), oil yield (dt/ha), seed oil content (% volume per seed dry weight), content of total seed glucosinolate (GSL; μmol/g seed), seedling emergence (visual observation ranging from a minimum value of 1 to maximum 9), lodging resistance (visual observation ranging from a minimum value of 1 to maximum 9) and days to onset of flowering (DTF; measured as number of days from sowing until 50% flowering plants per plot).Based on the observed population stratification in my dataset, I devised three scenarios for the genomic prediction. First, I considered the whole population, including all the pollinators (475) and then across two main clusters independently. In the whole population scenario, the highest prediction accuracy was achieved for seed oil content (rGPA = 0.81) and lowest for the least heritable trait, seedling emergence (rGPA = 0.29). No uniform improvement was seen in genomic prediction accuracies across individual clusters. The results of my study, however, suggest that prediction of testcross performance in hybrid spring-type canola breeding, where molecular variants are used across the whole genome, could be an efficient and cost-effective breeding approach to improve this important allopolyploid species.