Mapping loci controlling fatty acid profiles,oil and protein content by genome-wide association study in Brassica napus

Minqing Tng,Yunyun Zhng,Yueying Liu,Choo Tong,c,Xiohui Cheng,Wei Zhu,Ziyun Li,Junyn Hung,c,*,Shengyi Liu,c

a The Key Laboratory of Biology and Genetic Improvement of Oil Crops,The Ministry of Agriculture,Oil Crops Research Institute of Chinese Academy of Agricultural Sciences,Wuhan 430062,Hubei,China

b National Key Laboratory of Crop Genetic Improvement,National Center of Crop Molecular Breeding Technology,National Center of Oil Crop Improvement(Wuhan),Huazhong Agricultural University,Wuhan 430070,Hubei,China

c Hubei Collaborative Innovation Center for Green Transformation of Bio-Resources,Faculty of Life Science,Hubei University,Wuhan 430062,Hubei,China

Keywords:Brassica napus Oil content Fatty acid profiles Association study Haplotypes

A B S T R A C T Optimizing the profile and quantity of fatty acids in rapeseed(Brassica napus L.)is critical for maximizing the value of edible oil and biodiesel.However,selection of these complex seed quality traits is difficult before haplotypes controlling their contents are identified.To efficiently identify genetic loci influencing these traits and underlying candidate genes and networks,we performed a genome-wide association study(GWAS)of eight seed quality traits(oil and protein content,palmitic,stearic,oleic,linoleic,eicosenoic and erucic acids content).The GWAS population comprised 370 diverse accessions,which were phenotyped in five environments and genotyped using 60K SNP arrays.The results indicated that oil and protein contents generally showed negative correlations,while fatty acid contents showed positive or negative correlations,with palmitic and erucic acid contents directly affecting oil content.Seven SNPs on five chromosomes were associated with both seed oil and protein content,and five genes orthologous to genes in Arabidopsis thaliana were predicted as candidates.From resequencing data,besides known haplotypes in BnaA.FAE1.a and BnaC.FAE1.a,three accessions harboring a new haplotype conferring moderate erucic acid content were identified.Interestingly,in a haplotype block,one haplotype was associated with high palmitic acid content and low oil content,while the others showed the reverse effects.This finding was consistent with a negative correlation between palmitic acid and oil contents,suggesting historical selection for high oil content.The identification by this study of genetic variation and complex correlations of eight seed quality traits may be beneficial for crop selection strategies.

1.Introduction

Oilseed rape or rapeseed(Brassica napus,AACC,2n=38),one of the three major oil crops worldwide,is characterized by high oil and protein content[1].Itis a recentallotetraploid species derived from Brassica rapa[2](AA,2n=20)and Brassica oleracea[3](CC,2n=18).Owing to the importance of edible vegetable oil,protein feed,and potential energy generation,oilseed rape is assuming increasing importance worldwide,especially in view of the eventual exhaustion of petroleum resources and the urgency of environmental protection[4].The seed of oilseed rape contains mainly protein and oilconsisting offatty acids including palmitic(C16:0),stearic(C18:0),oleic(C18:1),linoleic(C18:2),eicosenoic(C20:1),and erucic(C22:1)acids[5].The improvement of these seed quality traits is a major goal of rapeseed breeding.However,owing to the complexity of the rapeseed genome and of seed biological processes,the genetics underlying seed quality traits are poorly understood.With the rapid development of biotechnology and high-throughput genotyping technology,the combination of genetic mapping with association mapping has now brought unprecedented possibilities for the development of superior crop cultivars with desirable traits[6].

The synthesis pathway of plant oil is a complex biochemical metabolic process,but the main biochemical metabolic reactions and corresponding enzymes in the biosynthesis of fatty acids have been extensively studied in Brassica species[7].Protein content of seeds was reported to be negatively correlated with oil content in oil crops[8].Quantitative trait locus(QTL)analyses indicate that oil and protein contents are controlled by many genes with additive and epistatic effects[9].For edible vegetable oil production,high contents of oil and C18:1 and low contents of C22:1 are the major goals of rapeseed breeding,while high protein content for feed is also desired.Genetic mapping has been employed to detect QTL in rapeseed,such as for oil content[10],silique length[11],and biomass and yield[12].Two major loci with additive effect that control mainly the synthesis of C22:1 in rapeseeds were found in the early 1960s[13],and were located on the respective linkage groups A08 and C03 by linkage mapping[14].Subsequently,two paralogous genes,BnaA.FAE1.a and BnaC.FAE1.a located at the locus controlling the synthesis of C22:1 were identified as the key factor determining seed oil quality[15].These two genes were also identified by associative transcriptomics[4]and association[16].A single nucleotide polymorphism(SNP)and two deletions were discovered in BnaA.FAE1.a and BnaC.FAE1.a,respectively,and their combinations were associated with different contents of C22:1[17].The elite alleles identified by association mapping for these traits have rarely been reported,to our knowledge,and an overview of networks involved in genetic control of fatty acid profile,oil content,and protein content is lacking.It is thus desirable to identify genetic variation of seed quality traits and the molecular mechanisms that determine seed quality for optimization of the energyefficient production of oil and protein in oil crops.

Genome-wide association studies(GWAS)can associate a large number of historical recombination events or genetic markers with target phenotypes in natural populations,thus facilitating elucidation of the genetic architectures of complex traits and the further discovery of elite alleles and haplotypes that could be applied to marker assisted selection(MAS)and pyramiding breeding.With rapid development in genomics and dramatic decrease in the cost of genotyping technology,GWAS has emerged as a powerful approach for identifying candidate genes underlying complex traits[4].Combining transcriptomics,metabolomics,and dynamic phenotypes allows GWAS to effectively identity natural variation and the genetic basis of metabolisms and phenotypes[18-20].

In this study,a panel of 370 rapeseed accessions was genotyped with the 60K Brassica Infinium SNP array.Data for eight quality traits(C16:0,C18:0,C18:1,C18:2,C20:1,C22:1,oil content and protein content)in B.napus seeds were collected in five environments.The aims of the study were(1)to identify SNPs significantly associated with traits by GWAS and investigate the genetic variation of candidate loci(or genes),(2)to identify the genetic mechanisms and relationships of these traits and construct networks genetic control of fatty acid profile,oil content,and protein content,and(3)to dissect a potential novel haplotype in BnaA.FAE1.a and BnaC.FAE1.a controlling fatty acid profiles.

2.Materials and methods

2.1.Plant materials,field experiments,and phenotyping

A worldwide collection consisting of 370 Brassica napus accessions(Table S1)was used for GWAS.The population phenotypic data were collected from field experiments of three years(2013-2015)in Wuhan,Hubei province and two years(2013 and 2015)in Yangzhou,Jiangsu province,China.The trial management followed standard breeding field protocols.Every line was planted in a 2 m2plot and field tests followed a randomized design with three replicates.Seed fatty acid composition,oil content,and protein content were used for quality evaluation.Open-pollinated seeds were harvested from 10 individual plants of each plot at maturity,and eight quality traits(C16:0,C18:0,C18:1,C18:2,C20:1,and C22:1,oil content and protein content)were measured for each plant in desiccated seeds with Foss NIR Systems 5000 Near-Infrared Reflectance Spectroscopy(NIRS)[21].

2.2.SNP genotyping,quality control,and location

Genomic DNA was extracted from young leaves of 370 selfpollinated lines at the seedling stage using the modified CTAB method[22].The DNA was hybridized to a 60K Illumina Infinium HD Assay SNP array(Illumina Inc.,San Diego,CA,USA)to obtain genotypes for 52,157 SNPs in each line.To guarantee the quality of SNP genotyping,in addition to biological and technological repeats,the SNPs were filtered based on call frequency≥85%,minor allele frequency≥5%,cluster separation scores≥0.4,heterozygosity≤15%,and unique physical position in the B.napus‘Darmor-bzh'reference genome(version 4.1)[1].

2.3.Statistical analysis of phenotype

As quality traits of natural populations are complex,the traits were evaluated in multiple environments and years with three replications.Best linear unbiased prediction(BLUP)was used in linear mixed models for the estimation of random effects.A linear model with random effects(replications,locations,and years)as variance components was fitted with the lme4 R package(www.eXtension.org/pages/61006),and the summary function was then used to calculate the estimate of 370 lines and the ranef function to calculate the BLUP value of each line.The frequency distributions of the eight seed traits were constructed with the hist function in R using a breaks parameter of 20.The broad-sense heritability(H2)of each trait in the natural population was calculated according to the result of the summary function for BLUP.Pairwise correlation coefficients for all traits were calculated with SAS 9.1.3(SAS Institute,Cary,NC,USA).Path analysis by multiple linear regression was performed with DPS software[23]using contents of C16:0,C18:0,C18:1,C18:2,C20:1,and C22:1 as independent variables and the oil content as a dependent variable.

2.4.Population structure,genetic relatedness analysis,and genome-wide association analysis

The population structure(Q)and genetic relationship(K)could produce false positive markers.Thus,it was necessary to take population and kinship into account when performing GWAS in TASSEL[31].The Q+K model(Q was a score matrix for each individual line,and K was calculated to estimate the pairwise relatedness between individuals)in a MLM was reported[16,32]to be the optimal model for GWAS.A subset of one third(7286)of the high quality 21,856 SNPs remaining after quality control was used to infer population structure with STRUCTURE 2.3.4[24].Five independent runs were performed with a K value varying from 1 to 10,and the numbers of burn-in and Markov chain Monte Carlo(MCMC)steps were both 10,000 under the admixture model.The△K results were used to estimate the number of subpopulations[25].Pairwise kinship was calculated with TASSEL v5.0[26]with the “scaled IBS”method,negative values were set to 0.GWAS was performed with the mixed linear model(MLM)using the Q+K model in TASSEL(significance level of Bonferroni-corrected threshold of P＜4.58×10-7,or 0.01/21856;-lg P=6.34).The results of GWAS were represented in a Manhattan plot using the R package CMplot of(https://github.com/YinLiLin/R-CMplot).The network of correlated traits and SNPs was displayed with Cytoscape 3.2.1[27].

2.5.Nucleotide-diversity analyses

PowerMarker[28]was used for basic statistical analysis and calculation of polymorphism information content(PIC)of the SNPs.SNPs significantly associated with C22:1 were extracted to estimate the linkage disequilibrium(LD)level to compare LDs among chromosomes and perform phylogenetic analysis.Selective sweep analysis was performed with DnaSPv5[29].The targeted genes BnaA.FAE1.a and BnaC.FAE1.a were re-sequenced in 324 accessions showing wide genetic variation to identify SNPs and InDels.

3.Results

3.1.Statistical analysis of phenotype and genotype

Eight quality traits in a natural population comprising 370 accessions were measured in three replications in five environments.Phenotypic diversities are described in Fig.S1 and Table 1.The mean broad-sense heritability was 0.91.C20:1 and C22:1 showed the highest heritability(0.98),and protein content the lowest heritability(0.76),still a high value in typical breeding studies.All the traits were consistent across environments and years.Evaluation of the pairwise phenotypic correlations(Table S2)showed highly significant(P＜0.01)correlations.C22:1 showed significant positive correlations with C20:1 and protein content,particularly with C20:1(r=0.98);in contrast,C22:1 showed significant negative(r=-0.99)correlations with C16:0,C18:0,C18:1,and C18:2,particularly with C18:1.Oil content showed relatively weak correlations with other traits,showing the highest correlation with protein content(r=-0.58).

Path analysis was performed for six fatty acid components(Table 2).The direct path coefficients of C16:0,C18:0,C18:1,C20:1,and C22:1 were negative on oil content.The corresponding direct path coefficient was-1.60 and the correlation coefficient was-0.23 for C16:0,indicating that C16:0 is important for oil content.Although C22:1 showed the highest direct path coefficient(-2.11),the correlation coefficient was 0.08(P=0.14).These results indicated that reducing C16:0 content is an efficient approach to increasing oil content in Brassica napus breeding.

The Brassica 60K SNP array with 52,157 SNPs was used to genotype the panel of 370 B.napus accessions.The source sequences of SNPs were blasted against the B.napus‘Darmorbzh'reference genome sequence[1]to identify their unique chromosomes and physical positions.High-quality 21,856 SNPs unique positions in reference genome remained(Table S3).Statistical analysis of these SNPs(Table S4)showed that chromosome A02 contained the fewest SNPs(458)with a SNP density of one SNP per 54.13 kb,and C04 the most SNPs(2554)with a SNP density of one SNP per 19.16 kb.The means of SNP heterozygosity and availability were 2.19%and 99.06%respectively,showing the high quality of the SNP dataset.Except for chromosomes C07 and C09,the PIC values of all chromosomes were greater 0.25.Chromosome A10 showed the highest PIC(0.32)and chromosome C09(0.23)the lowest.The mean PIC value of the full genome was 0.28,a value corresponding to moderate DNA marker polymorphism[30].

3.2.Population structure and linkage disequilibrium

The population structure of the association panel was estimated with STRUCTURE using 7286 SNPs uniformly distributed among chromosomes.Clustering inference performed with possible clusters showed an increasing of likelihood from 1 to 10(Fig.S2-A),and the highest ΔK value was observed at K=3(Fig.S2-B),suggesting that the 370 accessions could be assigned into three subgroups[25].Principal component analysis showed that the first two principal components explained 9.45%and 7.51%of molecularvariance.A three-dimensional diagram of the first three principal components(Fig.S2-C)also indicated that the panel could be divided into 3 subgroups.The blue and green groups comprised mainly commercial species and germplasm accessions,respectively,and accessions in the red group were developed from Zhongshuang 9 in our laboratory.The average relative kinship between any two accessions was 0.1077 and a total of 58.04%of the kinship coefficients were 0(Fig.S2-D),indicating that most accessions in the population had no or weak kinship.

Table 1-Statistical analysis a of eight seed quality traits.

LD in A subgenome,C subgenome and the full genome were analyzed using the SNP data in the population.The LD decay rate was measured as the chromosomal distance at which the average pairwise correlation coefficient(r2)dropped to 0.1.The genome-wide LD decay rate of 0.1 for the full genome was estimated as～240 kb,and those for the A and C subgenomes were estimated as ～185 and ～275 kb,respectively,but LD decay tended to uniformity at～375 kb(Fig.S3).The differences between the two subgenomes might have accumulated from a relatively long history of strong selection of large chromosome regions associated with important traits conferred by C-genome QTL[33].

3.3.Associated loci for contents of oil and protein

The results of GWAS showed that oil and protein contents were both associated with seven SNPs,with P value＜4.58×10-7(Fig.1 and Table S5).Eleven SNPs significantly associated with oil content were identified on five different chromosomes,while 10 SNPs located on the same five chromosomes were significantly associated with protein content.Seven of the ten SNPs were detected for both two traits and were distributed on chromosomes A04,A07,C01,C02,and C03.Five of them were most significant SNPs in their chromosome,revealing that there were not only significant(r=-0.58,P＜0.01)phenotypic associations but also genetic correlations between oil and protein contents.

As a comparison of these QTL with other previously reported QTL for oil content,two SNPs(Bn-scaff_18702_1-p589589 and Bn-scaff_20735_1-p42779)were previously reported(QTL OILC2-3[34]).A candidate gene BnaA04g17620D was found 103.2 kb from Bn-A04-p13738279,which encoded alpha/beta-hydrolase superfamily protein content. Its orthologous gene AT2G30550 encodes a lipase that has hydrolase,phosphatidylcholine 1-acylhydrolase,and triglyceride lipase activities in Arabidopsis[35].Two novel loci for oil content and some candidate genes,BnaA07g05070D and BnaC02g21440D,were located 82.17 kb from Bn-A07-p3583161 on A07 and 102.4 kb from Bn-scaff_20735_1-p42779 on C02,might be important for oil biosynthesis.The Arabidopsis ortholog of BnaA07g05070D(AT3G25110)is involved in fatty acid biosynthesis[36]and that of BnaC02g21440D(AT1G72520)is associated with lipid oxidation and oxylipin biosynthesis[35].For protein content loci,only 5.8 kb from Bn-A07-p3583161,a candidate gene BnaA07g05230D was found,whose orthologous gene AT3G24800 has the function ofproteolysis[37].Another candidate gene BnaC01g22960D was found 9.5 kb from the peak SNP in C01.Its orthologous gene AT4G25000 is involved in starch mobilization and the mutants were defective in alpha-amylase activity in Arabidopsis[38].

Table 2-Analysis of oil content path coefficient of six fatty acid components.

Fig.1-Manhattan plot for eight seed quality traits.Contents of C16:0,C18:0,C18:1,C18:2,C20:1,C22:1,oil and protein a displayed from inside to outside.Only chromosomes with significant SNPs are shown.Broken red circles show the Bonferronicorrected significance threshold,-lg(P)=6.34,for each trait.Red dots above each red circle indicate significantly associated SNPs with R2 values at the peaks indicated in parentheses.Significant SNPs associated with more than one trait are shown in wedge-shaped boxes.

3.4.Associated loci for contents of fatty acid components and their pathway

A total of 25 SNPs significantly associated with C22:1 content were detected by GWAS in two regions(8.54 to 11.04 Mb of A08 and 56.05 to 56.47 Mb of C03,respectively)(Fig.2-A and Table S5).Two peaks,Bn-A08-p12814556 on A08 and Bn-A08-p12660208 on C03,explained respectively 24.85%and 13.38%of total phenotypic variance.The two peaks were respectively 386.1 and 240.3 kb away from the key genes BnaA.FAE1.a and BnaC.FAE1.a,demonstrating that the association genetics approach was effective and precise for our natural population.Interestingly,the same peak SNPs were also detected on A08 for C16:0,C18:0,C18:1,C18:2,and C20:1,and on C03 for C16:0,C18:1,and C20:1(Fig.1 and Table S5).This finding,with path analysis,indicated that the fatty acid components were controlled by common genetic regions and enzyme systems in the fatty acid synthesis pathway in rapeseed.

LD is the nonrandom association of alleles at two or more loci descended from a single ancestral chromosome and is affected by allele frequency and recombination[39].Twentythree SNPs in the region from 8.54 to 11.04 Mb of A08 provided 253 significant pairwise values(P＜0.01),and the mean LD of the significant region was r2=0.76(Fig.S4-A),suggesting very strong LD.However,the average LD of A08 was just 0.24.Because only two SNPs were significantly associated with C22:1 on C03,the LD could not be calculated.Selective sweep is the reduction or elimination of variation among nucleotides in DNA near a mutation as the result of recent and strong positive natural selection[40].According to diversity analysis and Tajima test of accessions with a low C22:1 content(Fig.2-B,C),the mean diversity(Pi)and Tajima's D were 0.12 and-0.24 in the region from 10.18 Mb to 11.13 Mb on A08,values significantly different from the average of the whole chromosome(Pi=0.30 and Tajima's D=1.75).Similar results were obtained for the region from 55.44 Mb to 55.72 Mb on C03.These results indicated that the two genome regions were strongly selected during breeding,leading to the decrease in polymorphism.

Fig.2-Selective sweeps in the Bna.FAE1 regions.A:Manhattan plot of C22:1.The horizontal red line represents the Bonferronicorrected significance threshold-lg(P)=6.34 and the red dots above the red line the significantly associated SNPs.B and C indicate respectively the diversity(left axis)and Tajima's test(right axis)of chromosomes A08 and C03 for C22:1 content across accessions;D and E:Bna.FAE1 genes and their physical positions on A08 and C03.

During the 1980s and 1990s,low C22:1 rapeseed received great attention,and during the 2000s,it has been ubiquitously cultivated[5].The phylogenetic tree of the SNPs associated with C22:1(Fig.S4-B)showed that the natural population had three clear branches.Almost all accessions with a high or medium content of C22:1 clustered together(part 1),although a few accessions were close to those accessions with low C22:1(part 2)as revealed by the phylogenetic tree.Part 3 included almost all accessions with low C22:1,indicating that accessions with low C22:1 came from very few germplasm resources.As for the development history of the 370 accessions(Table S1),the majority of accessions(103 of 130)developed in the 2000s were accessions with low C22:1.However,most of those accessions(111 of 140)developed before the 1990s were the ones with a high content of C22:1,a finding consistent with the breeding history of low-C22:1 rapeseed cultivars.Taken together,these results suggested that the trait of C22:1 resulted mostly from modern breeding programs.

The resequencing of 324 of the 370 accessions showed that BnaA.FAE1.a had 1 SNP but no InDels and BnaC.FAE1.a had no SNPs but two Indels,with only one accession(EMS90)harboring a new AG deletion(Table S6)rather than an AA deletion as previously reported[17].The SNP mutant of BnaA.FAE1.a led its 282th Ser(the mean content of C22:1 was 11.49%)change to Phe(the mean content of C22:1 was 38.33%),and the 4-bp and 2-bp deletions of BnaC.FAE1.a resulted in low C22:1 content(10.32%and 11.89%)compared with the nodeletion accessions(26.39%).These results were similar to those of previous studies[15,41].In a previous report[17],the nucleotides C and T in the 845-bp position on BnaA.FAE1.a were designated haplotypes A08-H0 and A08-H1,respectively.The no-deletion of BnaC.FAE1.a was designated haplotype C03-H0,a 4-bp deletion in BnaC.FAE1.a and a 2-bp deletion in BnaC.FAE1.a were designated C03-H1 and C03-H2,respectively.A haplotype with two deletions in BnaC.FAE1.a was named C03-H3(Table 3).The different haplotype combinations were significantly lower than the wild type(combination of A08-H0 and C03-H0),in agreement with Wang[17].Three accessions(Luojingxuanxi,EMS-20,and EMS-92)contained an additional haplotype(a combination of A08-H0 and C03-H1),not found in previous studies,conferring moderate content(31.00%)of erucic acid.

A summary of the above results for GWAS and correlations of the eight quality traits(Fig.3),shows that the 8 traits notonly significant correlate in phenotypes but genetics.The contents of C16:0 and C22:1 had the largest effect on oil content(Table 2).A haplotype block,composed of Bnscaff_22466_1-p955103 and Bn-scaff_22466_1-p956140,was identified in which one haplotype(hap1:AG and AC)was associated with high C16:0 content and low oil content while the others(hap2:GG and CC or hap3:AA and AA)had the reverse effect,and an increase of 1.0%in C16:0 content caused by hap1 resulted in a decrease of 6.6%in oil content caused by hap2 or hap3,suggesting that the molecular markers enable selection for increased oil content by lowering C16:0 content.However,no loci explaining the negative relationship between C22:1 and oil content were identified.

Table 3--Erucic acid contents(%)in combinations of different Bna.FAE1 haplotypes in Brassica napus.

4.Discussion

4.1.Application of GWAS to Brassica napus

With the rapid development of new biotechnologies,especially the emergence and application of next-generation sequencing technologies,studies of genetic mechanisms have achieved great progress.Rapeseed contains two homologous but divergent subgenomes,A and C[1],hindering genomic research because of the complexity of the genome[16].GWAS is a powerful approach to identifying genes influencing traits in crops,and can provide valuable information for the improvement of crops and basic studies.The successful applications of associative transcriptomics[4]and the 60K SNP chip[16]indicated that GWAS is effective for detection of the genomic regions controlling traits in rapeseed.GWAS has been used to identify many genetic variants associated with seedling development traits[42],resistance to Sclerotinia stem rot[43],plant height[44,45],primary branch number[45],yield-related traits[32],and flowering time[46],but the detection of elite alleles for target traits has rarely been attempted.In the present study,we detected SNPs and haplotypes for fatty acid synthesis and analyzed genotypephenotype correlations using a 60K SNP chip and resequencing of targeted genes.Owing to the low density of SNPs(one SNP per 33.99 kb in the whole genome)and low level of phenotypic variation,the contributions of those significant SNPs for oil content and protein were too small to be identified.Therefore,it is necessary to expand the population size and perform whole genome resequencing to identify more molecular markers and candidate genes that control the target traits,which could provide valuable information for the construction of haplotype map and development of functional markers in rapeseed.

Fig.3-Graphical representation of the seed quality traits and associated SNPs detected by association analysis in Brassica napus.Gray lines correspond to the significant association of traits with SNPs.Colored lines connecting two traits correspond to Pearson correlations between trait phenotypic data.

We detected a region containing BnaA.FAE1.a gene on A08 for six intermediate metabolites(C16:0,18:0,C18:1,C18:2,C20:1,and C22:1)in the synthesis pathway of fatty acids.The results suggested that these metabolites might be regulated by BnaA.FAE1.a[7],resulting in the significant correlations and path correlations of the contents of the six fatty acids.Compared with GWAS based on conventional agronomic traits,metabolite-based GWAS could reveal gene functions more directly and effectively[20].It is possible to integrate the profiling data of transcriptomics,proteomics,metabolomics,and phenomics at different molecular levels(genome and transcriptome)to provide a more systemic perspective for relevant biological processes,and to more comprehensively dissect gene regulatory networks[47].As biological mechanism is powerful evidence to validate phenotypic mutation,further studies must combine biochemistry,cell biology,genetics,and functional analysis to reveal the biological association between phenotypes and genotypes[48].

4.2.The artificial selection of low C22:1 and co-location for quality traits in Brassica napus

Artificial selection is similar to natural selection,but the criteria for advantageous genes are set by human beings.It could influence the equilibrium of gene frequency by breaking the randomness,resulting in a change of gene frequency in a certain direction.LD analyses can provide insights into the history of both natural and artificial selection,and also can give valuable guidance to breeders who seek to diversify crop gene pools[33].In our study,the region on A08 significantly associated with the content of C22:1 showed higher LD,lower polymorphism and more negative Tajima's D than the mean of its chromosome(Figs.2-B,D,and S4-A).These results indicated that the significant region of low polymorphism could be a result of human intervention,consistent with the breeding history of low C22:1 content.

Significant correlation among phenotypes is a common phenomenon in crops.One reason for the negative correlation between oil and protein contents in the seed could be the competition for the same basic substrates in the biochemical pathways of oil and protein[49].Both oil and protein contents in the seed are quantitatively inherited traits determined by the interaction of multiple genes subjected to the interactions between genotype and environment in rapeseed[49]and soybean[50].When key genes of the fatty acid biosynthetic pathway were overexpressed in low seed storage protein content mutants,the seed oil content was increased,while the protein content was reduced[51].Future work is needed to identify candidate genes regulate the oil and/or protein biosynthetic pathways,and elucidate the precise molecular functions of these candidate genes.This work will help to identify pleiotropic loci with simultaneous effects on multiple traits.In the present study,we detected seven common SNPs for both oil and protein contents.We expect these SNPs to be effective and helpful for pyramiding multiple traits in rapeseed.These associated SNPs could be the source to obtain optimal composition in rapeseed breeding for different applications with maximal economic efficiency.

5.Conclusion

Using 21,856 SNPs with a genetically diverse set of 370 rapeseed germplasm accessions to perform GWAS,aiming to detect genome regions controlling eight seed quality traits.We dissected the genetic variation of these quality traits and the predicted candidate genes.We found phenotypic correlations between fatty acids controlled mainly by BnaA.FAE1.a.We identified SNPs associated with both oil and protein content.Using the resequencing data,we identified three lines contained a novel haplotype in Bna.FAE1 controlling moderate content of C22:1.These valuable results are of great significance in rapeseed production and improvement.With the rapid development of-omics studies and dramatic decrease in the cost of biotechnology,the combination of genomics,transcriptomics,proteomics,metabolomics,molecular biology,and other disciplines to identify the regulatory network of important traits will be the main direction in future rapeseed research.

Acknowledgments

This research was supported by the National Key Research and Development Program of China(2016YFD0101007,2016YFD0100305,2018YFD0200904),the National Natural Science Foundation of China (31471536,31770250),the Earmarked Fund for China Agriculture Research System(CARS-12)and the Agricultural Science and Technology Innovation Program(ASTIP)of Chinese Academy of Agricultural Sciences.

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2018.10.007.