Meng Wng,Grhm J.King,Lei Shi,Ruiyun Li,c,Yi Zhng,Xiohu Wng,Jinling Meng,Jinxing Tu,Jun Zou,*

a National Key Laboratory of Crop Genetic Improvement,Huazhong Agricultural University,Wuhan 430070,Hubei,China

b Southern Cross Plant Science,Southern Cross University,PO Box 157,Lismore,New South Wales 2480,Australia

c Key Laboratory of Information and Computing Science of Guizhou Province,Guizhou Normal University,Guiyang 550001,Guizhou,China

d College of Agriculture and Forestry Science,Linyi University,Linyi 276000,Shandong,China

Keywords:Allopolyploid Meiotic crossovers Brassica napus Historical introgression Reciprocal backcross population

ABSTRACT Variation in patterns of recombination in plant genomes provides information about species evolution,genetic diversity and crop improvement.We investigated meiotic crossovers generated in biparental segregating and reciprocal backcross populations of the allopolyploid genome of rapeseed(Brassica napus)(AACC,2n=38).A structured set of 1445 intercrossed lines was derived from two homozygous de novo genome-assembled parents that represented the major genetic clusters of semi-winter Chinese and winter European rapeseeds,and was used to increase QTL resolution and achieve genomic reciprocal introgression.A high-density genetic map constructed with 6161 genetic bins and anchored centromere regions was used to establish the pattern of recombination variation in each chromosome.Around 93%of the genome contained crossovers at a mean rate of 3.8 cM Mb?1,with the remaining 7% attributed to centromeres or low marker density.Recombination hotspots predominated in the A genome,including two-thirds of those associated with breeding introgression from B.rapa.Genetic background might affect recombination variation.Introgression of genetic diversity from European winter to Chinese semi-winter rapeseed showed an increase in crossover rate under the semi-winter environment.Evidence for an elevated recombination rate having historically contributed to selective trait improvement includes accumulation of favorable alleles for seed oil content on hotspots of chromosome A10.Conversely,strong artificial selection may affect recombination rate variation,as appears to be the case with a coldspot resulting from strong selection for glucosinolate alleles on A09.But the cold region would be promptly reactivated by crossing design indicated by the pedigree analysis.Knowledge of recombination hotspots and coldspots associated with QTL for 22 traits can guide selection strategies for introgression breeding between the two gene pools.These results and rich genomic resources broaden our understanding of recombination behavior in allopolyploids and may advance rapeseed genetic improvement.

1.Introduction

Meiotic recombination is a key biological process in plant sexual reproduction.It influences genome evolution,population genetic diversity,and crop improvement.The meiotic crossover rate(recombination rate,θ)not only varies among different species[1],but is also uneven within and between constituent chromosomes of a species,with recombination hotspots and coldspots[2,3].Allelic,transcriptional,and phenotypic variation is often associated with recombination hotspots as well as recombination within genes is also associated with recombination hotspots[4].Evidence identifying the chromosomal recombination landscape in plants has come mostly from diploid species such as Arabidopsis thaliana,and is relatively rare for species of more recent polyploid origin.Genome composition and complexity vary widely among crop plant species,resulting from diverse ancestral evolution and post-domestication breeding histories,with consequent variation in the distribution of recombination.Establishing recombination landscapes for a range of taxa,especially those having complex polyploid genome compositions,is likely to produce systematic and scalable information for understanding and manipulating recombination for breeding improvement or management of biodiversity.

Brassica napus(AACC,2n=38)is a young domesticated allopolyploid species formed～7500 years ago[5].It has a relatively complex genome arising from interspecific hybridization between B.rapa(AA,2n=20)and B.oleracea(CC,2n=18),each of which has a well-documented hexaploid ancestry[6].Although we cannot trace the first domestication of B.napus for human cultivation,it is clear that B.napus has also undergone several rounds of secondary domestication,contributing to its global prominence as a temperate oil crop(known as canola or rapeseed).This includes introgression of segments from the diploid A genome[7].Three ecogeographical crop types(winter,spring and semi-winter)of B.napus are recognized,classified according to distinctive plant habit,growing region,and requirements for vernalization[8].These correspond to three major genetic clusters of B.napus germplasm that display substantial population differentiation formed over its very short history of domestication and selection[5,7–9].For rapeseed genetic improvement,breeders commonly broaden the genetic diversity and use heterosis in hybrid development by reciprocal introgression of these different genetic clusters[10].

Several studies[11–13]have identified recombination rate variation in the genus Brassica.In 67 diverse B.napus doubled haploid(DH)lines,the A genome accumulated more crossovers(CO)and gene conversion events than the C genome[14].More recently[15],a comparison of diverse cultivars representing three B.napus seasonal crop types indicated higher θ in winter than in spring and semi-winter types.Investigations based on a diverse range of cultivars are likely to reflect cumulative historical artificial selection for recombination behavior in various chromosomal regions.However,in turn,recombination may be subject to other processes including strong artificial selection for specific post-domestication traits and introgression from diverse genetic backgrounds.Recombination behavior diverse cultivars contrasts with studies based on multiple generations of lineages derived from populations having a fixed genetic background[16].Advanced-backcross population designs have been used to characterize the similarities and differences in recombination between different genetic backgrounds,such as in Vitis rotundifolia[17].We expected that establishing a lineage-specific recombination landscape by tracking meiosis arising in a hybrid derived from representatives of distinct genetic clusters and subsequent reciprocal backcross lineages to provide genetic information on variations in parameters affecting recombination in the B.napus gene pool.

We developed biparental segregating and reciprocal backcross populations with well-defined pedigrees to achieve genomic reciprocal introgression and establish a detailed recombination map for specific B.napus lineages.The map reflects the distribution of meiotic crossovers detected within a structured set of intercrossed lines derived from two homozygous parents that represent the Chinese semi-winter and European winter genetic clusters of rapeseeds.The BnaTNDH population has previously been used as a reference for a range of genomic and genetic mapping studies in multiple environments(https://www.brassica.info/)including establishment of rich genotypes using various marker types[18,19],phenotypes of 22 traits evaluated in 19 environments and hundreds of consensus QTL[20].The reciprocal backcross populations were used for QTL resolution and cloning of genes controlling agronomic and quality traits.Reference genomes are available for both parental lines[7,9,21],along with genomic reconstruction of parental pedigrees[7,22].We aimed to use the recombination map to investigate variation associated with reciprocal introgression of genome segments into the two parental lines.This would enable us to characterize genome-wide recombination between the two major genetic clusters and its effects on trait improvement and introgression breeding based on quantifying identity-bydescent(IBD)inheritance.

2.Materials and methods

2.1.Plant materials and phenotyping

2.1.1.Founder parent lines

The two homozygous founder parental lines have been well described[7,18].Tapidor is a DH line derived from a European winter cultivar with double-low seed quality(low glucosinolate and low erucic acid content).Ningyou 7(NY7)is a DH line derived from a Chinese semi-winter cultivar with double-high seed quality(high glucosinolate and high erucic acid content)(Fig.1).The genome sequences of both lines have been de novo assembled[7,9,21].

2.1.2.Segregating and reciprocal backcross populations

Three populations were derived from an F1of a cross between Tapidor and NY7(Fig.1).The initial unselected segregating BnaTNDH_353 population consisted of 353 lines derived by microspore culture from an F1hybrid between a doubled haploid derived from Tapidor and NY7[7].In previous studies[20,23],we used a subset of this population(BnaTNDH_182)to identify 1904 consensus quantitative trait loci(QTL)accounting for 22 traits in up to 19 environments using a dense SNP-based genetic map.The other subsets of BnaTNDH populations have also been used to identify centromeres and stable DNA methylation-marked loci and a wide range of QTL[23,24].

To generate sets of backcross lines in which segments of each parental genome are substituted in the background of the other,recurrent reciprocal backcrosses were made(Fig.1).Pollen from the original parent Tapidor was backcrossed to each of 83 BnaTNDH lines,with the DH lines acting as the female recipient.Unselected single-seed descent was used to generate the BC lineage from each BnaTNDH line to BC2F1,which was sufficient(P＞0.999,in a binomial distribution)to ensure that the set of lines can introgress any donor genome segments.Given that additional segregation and crossovers occur in the generations between BC2F1and BC4F1,a single plant cannot represent all recombination events in progeny families derived from an original DH line.Our strategy was accordingly to sample a representative set of crossovers in the backcross progeny,where introgressed segments were successfully captured to cover the whole genome.This goal was achieved by expanding the population based on the ratio of increasing recurrent parent components according to 1/2(n+1)(where n represents the backcross generation),following a similar process reported for other crops,such as rice(Oryza sativa L.)[25].The unselected introgression process was repeated for each BC lineage to BC4F1,from which a BC4F2population was generated by selfing and named BnaN-BC4F2(the population with the introgression of chromosome segments from NY7)(Fig.1).An equivalent reciprocal BC process was performed with pollen from the original NY7 male parent used to pollinate each of 60 DH lines,28 of which overlapped with DH lines for generating the BnaN-BC4F2population.

In summary,116 DH lines as the female recipient were used to construct the two reciprocal backcross populations,and the crossovers in these DH lines covered 99.9% of the whole BnaTNDH_353 population(Fig.S1).The BnaN-BC4F2population with Tapidor as a recurrent parent consisted of 1011 lines derived from 83 DH pollen recipient lines,with five to 15 BC4F2per DH line.The reciprocal BnaT-BC4F2population with NY7 as a recurrent parent consisted of 956 lines derived from 60 DH pollen recipient lines,with six to 18 BC4F2per DH line(Fig.1).The reciprocal backcross populations were planted from Oct.2014 to May 2015 in Wuhan,Hubei province,China,and the flowering time and quality traits were investigated.

Fig.1.Construction of BnaTN populations and the pedigree of their parents Tapidor and NY7.The original founder parent Shengliyoucai(SL)in the Ningyou 7 lineage is one of the first popular B.napus cultivars in China,introduced from Japan in the 1930s.Chengduaiyoucai(CDA)is a B.rapa cultivar with early flowering developed in the 1950s.Ningyou 1(NY1)is a semi-winter cultivar bred in Jiangsu province of China in the 1950s.Chuanyou 2(CY2)is a semi-winter cultivar bred in Sichuan province of China in the 1950s.Liho and Bronowski are the first and widely used donors of low erucic acid and glucosinolate for B.napus identified in Germany and Poland in the 1970s.Regent is a double-low cultivar bred in Canada in 1977,and Bienvenu is a high-yielding and high seed glucosinolate cultivar bred in France in 1982.The original parents have been described in previous studies[7,22]and their genotypes determined by the Brassica 60K-SNP and 100×genome sequence are available.Pedigree outlining the process used to generate the BnaT-BC4F2 and BnaT-BC4F2 populations.In the reciprocal backcrossing,Tapidor and NY7 were used as pollen donors.

2.2.Genotyping the BnaTNDH,BnaN-BC4F2,and BnaT-BC4F2 populations

The Illumina Infinium Brassica 60K SNP BeadChip array was used to generate genotype data for the complete BnaTNDH_353 population,of which data for 196 lines had previously been reported[23].Single nucleotide polymorphism(SNP)markers with sequences information were aligned to the NY7 reference genome[7]with BLASTn as described previously[23].For the BnaN-BC4F2population,we selected a subset of 769 lines from 1011 lines,some of which were removed during the study,representing lineages derived from 83 DH parents,with 1 to 12 lines per DH line.These lines were genotyped by the double-digested restriction-site associated DNA(ddRAD)sequencing method on an Illumina HiSeq 3000 system at national key laboratory of crop genetic improvement of Huazhong Agricultural University in China with pairedend sequencing of 150-bp inserts,with a 0.1×mean depth of coverage.For the BnaT-BC4F2population,323 lines were randomly sampled from the representative pedigree origins(involving 60 DH parents,with 1–8 lines corresponding to each DH line).Each line was resequenced to 5×genome coverage on an Illumina HiSeq X-Ten system.

Raw paired-end reads from ddRAD and resequencing were first filtered using NGS QC toolkit v2.3.315(https://hpc.ilri.cgiar.org/_export/xhtml/ngsqctoolkit-software).The cutoff value for the Phred quality score(an integer value representing the estimated the accuracy of the sequencing bases)was set to 20,and the percentage of read length that met the given quality was 70.Clean reads for each line were mapped to the NY7 reference genome using BOWTIE2 2.3.4 software[26]with default settings.PCR duplicates were removed with SAMTools software(https://samtools.sourceforge.net/)and Picard(https://broadinstitute.github.io/picard/).SNPs and InDels were called with the Genome Analysis Toolkit(GATK 3.8)[27].SNPs with＞10% missing genotypes and a minor-allele frequency＜5%were discarded.Genotype phasing was performed with Beagle 4.1[28].

2.3.Construction and integration of high-density genetic linkage maps

The high-quality markers remaining for each population were assigned to genetic bins to establish their redundancy.A Perl script was used to assign marker pairs with identical genotypes to genetic bins where they shared the same segregation pattern.For each bin a single representative marker with the highest quality was chosen to construct a genetic linkage map with JoinMap 5.0[29].The threshold for goodness of fit jump was set at≤5.0 with a recombination rate of＜0.4 and a minimum logarithm of odds score of 1.0.Markers with a χ2value of＞3.0 were excluded from all genetic groups[23].Recombination estimates were converted to centimorgans(cM)using the Kosambi function for map distance calculation.Double-crossover and recombination frequencies were checked with MapDisto 2.0[30].An integrated genetic map was constructed using Consensus Map Construction(CMP)in QTL IciMapping 4.2 software[31].In addition,the Spearman correlation coefficient were used to assess the level of collinearity between genetic and physical locations.

2.4.Anchoring positions of the centromeres on the genome

Centromere-specific markers were previously[24]identified using the BnaTNDH population.These were re-anchored on the newly constructed BnaTNDH genetic map and located in the reference NY7 and Tapidor genomes.RepeatMasker software[32]was used to verify the location of the centromeres according to specific centromere-repeat sequence characteristics.For chromosomes lacking centromere markers,centromere positions were determined based on alignment with centromere-specific sequence positions in the NY7 and Tapidor reference genomes.

2.5.Identification of recombination events,recombination rate,hotspots,and coldspots

A custom R script was used to identify recombination events.Considering the possible errors caused by genotyping and unlikely double crossovers,we chose only regions spanning at least four continued loci for the calculation of recombination events.Estimation of recombination rate θ(cM Mb-1)relies on accurate measurement of both physical and genetic distances in the genome.We accordingly discarded markers lacking a confirmed physical position[4,33].Recombination rates were estimated using MareyMap software[34].The recombination map was constructed and displayed with a 2-Mb window size calculated from an overlapping 1 Mb step using the‘‘Sliding window”method in MareyMap.In the(peri-)centromere regions,the segregation ratio of genotypes in the populations was calculated for assessing the recombination in these regions.Recombination hotspots were identified using a global permutation threshold as the cutoff above a significance level of 0.05.Coldspots were located in regions(＞1 Mb)devoid of recombination(0 cM Mb-1)[3,4].The genomic features of the hotspots and coldspots were identified based on the NY7 genome.SNP,InDel and gene densities were identified with a nonoverlapping sliding window of 100 kb used for density statistics.Gene density was calculated as the proportional base count occupancy in each 100 kb.The Advanced Circos from TBtools software[35]was used for visual presentation,along with SNP and InDel placement.

2.6.Anchoring QTL on the NY7 genome

All SNP sequences detected by the Illumina Infinium Brassica 60K SNP array used for construction of the BnaTNDH genetic linkage map were aligned to the NY7 genome using BLASTn with an E value＜10-10.Unique SNPs were assigned a physical position where they had a unique match,100% alignment and mismatch＜0.Unique SNP markers were used to determine the location of QTL detected in the BnaTNDH population[20].

3.Results

3.1.Genome-wide reciprocal introgression and construction of genetic maps from biparental segregating and reciprocal backcross populations

3.1.1.Genome-wide reciprocal introgression between semi-winter Chinese and winter European Brassica napus

The BnaTNDH,BnaT-BC4F2and BnaN-BC4F2populations were genotyped with respectively 23,470,122,005 and 18,425 highquality polymorphic markers.The mean marker densities were 2.6 100 kb-1(BnaTNDH),13.7 100 kb-1(BnaT-BC4F2)and 2.1 100 kb-1(BnaN-BC4F2).Although the density was higher for BnaTBC4F2than for the other two populations,marker distributions from the three populations on the NY7 genome were consistent,and the marker distributions also consistent with the known genetic diversity between the parents Tapidor and NY7(Fig.2).

The genotype data for the BnaT-BC4F2population indicated that the proportion of recurrent parent NY7 ranged from 72.0%to 99.7%(Table S1),with a mean of 1.8% of the genome regions covered by homozygous introgression segments from Tapidor and 7.8% of the genome covered by heterozygous segments.A mean of 1.6 introgressed segments with a mean length of 2.7 Mb were detected per chromosome per line,with a maximum of 13(on chromosome C07)and a minimum of zero(Fig.S1).For the BnaN-BC4F2population,the proportion of recurrent parent Tapidor ranged from 80.1%to 97.2%as expected(Table S1),with a mean of 2.8%of the genome regions covered by homozygous introgressed segments from the donor parent NY7,and 5.1% of the genome covered by heterozygous.A mean of 0.5 introgressed segments with a mean length of 8.0 Mb per chromosome per line was detected,with a maximum of 11(on chromosome C01)and a minimum of zero(Fig.S1).For the reciprocal backcross populations,the whole genome of the respective donor Tapidor and NY7 parent was covered in the introgression segment set(Fig.S1).

3.1.2.Construction of high-density genetic maps and the locations of centromeres

Genetic bin mapping was used for each population,with SNP markers from the BnaTNDH(23,470 markers),BnaT-BC4F2(13,304 markers)and BnaN-BC4F2(18,425 markers)populations independently allocated to 6610,13,304,and 8922 genetic bins.The resulting genetic maps contained 2513 genetic bin loci with 10,739 markers,2605 loci with 30,489 markers,and 2570 loci with 7995 markers(Tables 1,S2).High congruence between the three genetic maps and the NY7 physical map was observed,with Spearman correlations＞0.9 for most(15)chromosomes(Fig.S2;Table S3).The three BnaTN genetic maps were combined to form a high-density BnaTN-integrated genetic map(Table S2),which contained a total of 6161 informative genetic bins including 38,272 markers.This density was greater than that of any individual map.Congruence between the integrated map and the NY7 physical map for all chromosomes exceeded a coefficient of 0.9(Table S3).Compared with the BnaTNDH genetic maps,those of the reciprocal backcross populations showed higher genome coverage.The genome coverage in linkage group A04 was low(88.2%)in the BnaTNDH,but was greater in the BnaT-BC4F2(96.9%)and BnaN-BC4F2(99.2%)genetic maps.Integration of the maps further increased the coverage.The genome coverage of linkage group C07 was 99.5% in the integrated map but only 94.1%,95.6% and 97.2% in the three BnaTN genetic maps.

Fig.2.The distribution of markers and recombination rate of the BnaTN populations on the NY7 genome.I,SNP and InDel density between Tapidor and NY7;II,Polymorphic marker density of the three BnaTN populations.Black,brown,and green lines represent BnaTNDH,BnaT-BC4F2 and BnaN-BC4F2 population,respectively;III,θ of BnaTNDH population;IV,θ of BnaT-BC4F2 population;V,θ of BnaN-BC4F2 population;VI,θ of BnaTN-integrated genetic map.Blue,red,yellow and cyan points represent recombination hotspots identified in BnaTNDH,BnaT-BC4F2,BnaN-BC4F2 population and BnaTN-integrated genetic map.

Centromere locations were determined for each chromosome(Table S4).The centromeres of 17 chromosomes were identified based on the criteria above.The centromeres of chromosomes A02 and C05 were determined solely from the centromerespecific repeat sequences of the parental genome assemblies.Centromeric regions showed low marker densities,except for those on chromosomes C01 and C04(Fig.2;Table S4).Centromere regions on chromosomes A01–A10,C02,C04,and C09 were shortened by a mean of 2.79 Mb(Table S4).Apparent centromere length varied widely,from 0.13 Mb(A01)to 6.31 Mb(C01),with a mean length of 2.52 Mb for all 19 centromeres.Overall,the centromeres were markedly shorter(1.56 Mb)for A genome than for C genome chromosomes(3.59 Mb)(Table S4).These results of high-quality linkage data and the locations of the centromeric regions were used for the subsequent analysis of genome-wide recombination variation.

3.2.A composite map of recombination variation between semi-winter Chinese and winter European Brassica napus

3.2.1.Crossover representation in relation to the rapeseed genome

The integrated high-density genetic map was used to identify genome-wide θ variation,using a 2-Mb window size with a step size of 1 Mb.Genetic loci from the integrated genetic map were spaced at a mean density of 6.9 loci Mb-1(every 145 kb)in the assembled NY7 genome,sufficient for estimates of θ.To avoid the detection of false positives,a region＞1 Mb larger than the step size was used to determine the presence of recombination coldspots devoid of recombination.A mean genome-wide θ of 3.23 cM Mb-1(respectively 4.32 and 2.59 cM Mb-1for the A and C genomes)was observed(Table S5),with 16 conserved(detected in all three populations)recombination coldspots,including four in the A and 12 in the C genome(Table S6).Approximately 56%of the conserved coldspots were located in pericentromeric regions(A02,A03,A08,A10,C02,C07 and C08),and the remainder were associated with low gene and SNP or InDel densities.Conserved coldspots covered 7.4% of the NY7 genome,so that 92.6% of the whole genome contained crossovers(94.0% of the A and 91.8% of the C genome)with a mean rate of 3.8 cM Mb-1.

3.2.2.Hotspots predominated in the a genome,including a subset associated with breeding introgression

A recombination variation map of the BnaTN populations enabled us to characterize recombination hotspot variation in more detail,indicated by a θ significantly greater than thepopulation-specific thresholds(Table S6)over regions＞0.1 Mb.A total of 121 recombination hotspots(70 on the A genome and 51 on the C genome)were identified,covering 16.6% of the NY7 genome(Fig.2;Tables 2,S6).Around 40%of hotspots represented twothirds of those on the A genome and a third of the total,located in regions of the A genome corresponding to the introgression of the founder B.rapa parent‘Chengduaiyoucai’(CDA)into NY7(Fig.S3).Only seven overlapping hotspots were common to the three populations,with six corresponding to 2.2% of the A genome(A01,A08 and A10)and one to 0.2% of the C genome(C05)(Fig.2;Tables 2,S6).Of these conserved hotspots,five were located in regions of known breeding introgression of B.rapa(Fig.S3).As shown in the same DH line(DH167)and its derived reciprocal backcross progeny,four hotspots on four chromosomes involved with the introgression of CDA that was identified by tracing IBD inheritance(Fig.3).

Table 1 Summary statistics of the genetic maps constructed with the BnaTN populations.

The hotspots were compared with those identified in other published B.napus populations(Table S6)[36–38].Fifty-six hotspots overlapped those in more than two populations,with 44 on the A and 12 on the C genome.For example,two hotspots on A01 and A10 overlapped those in other published populations(including synthetic rapeseed-derived populations)[36–39].Sixty-five hotspots were unique to the BnaTN populations,of which 26 were on the A genome(37% of the hotspots on the A genome)and 39 on the C genome(76% of the hotspots on the C genome).This finding suggests that hotspots are more prevalent in the A than in the C genome.

3.2.3.Genetic background effects on recombination variation

There was no significant difference in θ between the BnaTNDH(3.61 cM Mb-1)and BnaT-BC4F2(3.35)maps,but both were significantly higher than in the BnaN-BC4F2(2.53)(Table S5).Likewise,the mean recombination events per line were similar in the BnaTNDH population(22)and BnaT-BC4F2(23)and both significantly more than the 12 detected in the BnaN-BC4F2population(Table S7),whereas the maximum events per chromosome were similar(11–13)in all three populations(Table S7).We investigated the contribution of common DH lines to the detection of recombination events within BC4F2lines.On average,more recombination events(18)were detected in the BnaT-BC4F2lines than in the Bna-N-BC4F2population(11)(Table S7).Whereas 25 recombination events were detected in DH167,means of 21(BnaT-BC4F2)and 9(BnaN-BC4F2)were detected in the corresponding derived BC4F2progenies(Fig.3;Table S7).The higher θ and greater number of recombination events in BnaTNDH and BnaT-BC4F2may be partially due to the historical introgression of A genome segments from B.rapa into the NY7 parent(Fig.S3).

Considering the possibility of the influence of population marker density on recombination statistics,we randomly sampled a subset of 5000 markers from all three populations to verify the results that higher θ and greater number of recombination events in BnaTNDH and BnaT-BC4F2.This test showed that the mean number of recombination events per line was consistent with those calculated from the full marker dataset(Table S7).Although there were differences in genotyping methods between the three populations,the distribution of markers on the NY7 genome was substantially the same for the three populations(Fig.2).This finding suggests that differences in genotyping methods did not strongly influence estimates of θ.

These findings suggest that recombination hotspot locations are mostly not conserved in the genome and are sensitive to genetic background.The number of hotspots on each of the A and C genomes was similar in the BnaTNDH(34 on A and 14 on C)and BnaN-BC4F2population(31 and 16),but differed from the BnaTBC4F2population(18 and 21)(Table 2).Hotspots in the reciprocal backcross populations covered 7.5%and 13.8%of the A genome and 4.1%and 3.6%of the C genome,whereas they covered 16.5%of the A genome and 3.3% of the C genome in the BnaTNDH population(Table 2).Hotspots that overlap between BnaTNDH and the reciprocal backcross populations covered 3.5% and 4.8% of the Agenome,and 0.6% and 3.3% of the C genome,whereas hotspots overlapping between the reciprocal backcross populations covered less(3.3%of the A and 0.5%of the C genome)(Table S6).These findings are consistent with an effect of genetic background on recombination hotspot distribution,with reduced crossovers where Tapidor is the predominant genetic background compared with NY7,which may be partially due to the historical introgression of the diploid A genome in NY7.

Table 2 Comparison of recombination hotspots among the BnaTN populations.

Fig.3.Genetic structure and IBD inheritance of a single DH line and its derived reciprocal backcross progeny.We take DH167 from the BnaTNDH population as an example.TBC4F2_L31,T-BC4F2_L179(BnaT-BC4F2),N-BC4F2_L79,and N-BC4F2_L498(BnaN-BC4F2)are DH167(a TNDH line)derived BC4F2 progeny.In the DH line and BC4F2 progeny,the light blue represents a homozygous introgressed segment from NY7.The orange color represents homozygous introgressed segments from Tapidor,and the blue color heterozygousis segments.Small blue,red and yellow circles represent recombination hotspots identified in respectively the BnaTNDH,BnaT-BC4F2 and BnaN-BC4F2 populations.Yellow arrows indicate introgression from B.rapa‘Chengduaiyoucai’(CDA).

3.3.The effect of recombination on QTL resolution and consequent influence by artificial selection

3.3.1.Use of backcross populations increases resolution of QTL

In this study,high-resolution genotype data for progeny of a biparental cross were available,and multiple phenotypic traits were investigated in reciprocal backcross populations.Taking the flowering time and two seed quality traits as examples,we could increase QTL resolution by identifying overlapping segments in lines from the reciprocal introgression populations having opposing extreme phenotypes.Thus,comparison of the genome-wide introgressed segments in lines with extremely early flowering(much earlier than that of the recurrent parent Tapidor)and later flowering time in the BnaN-BC4F2population,we identified a 237-kb segment located between markers MN8588 and MN8600 on chromosome A10.This segment contributes to the control of flowering time(Fig.4),and lies in a 1.2 Mb region previously identified in the BnaTNDH population.Among the 47 genes located in this segment,we found the gene BnFLC.A10,known to regulate flowering time and seasonal crop type(Table S8).

By selecting lines from the BnaT-BC4F2population with high and low seed erucic acid content and those with high and low glucosinolate content,we identified a 941-kb introgressed segment between markers MT47207 and MT47228 on A08 that controls erucic acid content and a 1017-kb segment between markers MT84535 and MT84622 on chromosome C02 that controls glucosinolate content(Fig.4).These represent a 5-fold(from 4.9 Mb)and 2.8-fold(from 2.8 Mb)increase in resolution compared with QTL previously detected in the BnaTNDH population.There were 118 genes in the A08 segment,including the gene FAE1 controlling erucic acid content.In the C02 segment,there were 125 genes,including two candidates,BnaC02.MYB28 and BnaC02.MYB34(Table S8).Compared with the BnaTNDH population,using the backcross populations cgreatly shortened QTL intervals,indicating the high resolution contributed by recombination occurring on the corresponding genome regions in backcross populations.Although we did not clone the alleles of the genes underlying the QTL effect in these regions,the populations and the overlapping-segment analysis demonstrate the power of unselected recurrent BC introgressions for resolving initial QTL.

3.3.2.Recombination variation and trait improvement affect each other

A total of 1905 consensus QTL for 22 traits(including yield and yield-related traits,developmental traits,and seed quality traits)have previously been detected on linkage maps derived from the BnaTNDH population,of which 1749(92%)have been anchored to 91.9% of the assembled NY7 genome(Fig.S3).Of these,500 QTL(26%)are located in recombination hotspots covering 8% of the NY7 genome.These QTL had a mean confidence interval of 1.00 Mb,over seven times shorter than the mean confidence interval of 7.14 Mb for the 208 QTL located within coldspots that correspond to 35% of the NY7 genome(Table S9).

Given sufficient positive selection,genomic regions with high θ can contribute to the rapid accumulation of favorable alleles and increase the efficiency of genetic improvement.Among the QTL located in hotspot regions,we identified three known SNP loci(Bn-A10-p16326058,Bn-A10-p16133231,and Bn-A10-p15924755)associated with seed oil content in chromosome A10 hotspots.These loci with favorable alleles have also accumulated in ZS11(an elite high oil-content cultivar released in the 2000s in China).This finding suggests that favorable alleles associated with seed oil content located in A10 recombination hotspots can accumulate readily in modern B.napus cultivars.

Conversely,artificial selection can also alter the recombination rate of genomic regions associated with a target trait by fixation of favorable alleles.A 3–5 Mb region on chromosome A09 includes a cluster of glucosinolate QTL which could explain mean 17%of phenotypic variance(Table S9),where genes involved in glucosinolate biosynthesis and regulation have been located.The IBD inheritance pattern for the Tapidor pedigree indicates that this segment of A09 was inherited intact in Tapidor from the founder parent Bronowski(the primary donor of low glucosinolate in modern canola-type rapeseed)by consistent selection for the low-glucosinolate trait in the parents(Figs.1,5).Although there was no selection for low glucosinolate in any of the three BnaTN populations,we observed opposing recombination rates in this region,with a recombination coldspot detected in the BnaT-BC4F2population that contained the introgressed segment from Tapidor.In contrast,a recombination hotspot was observed in this region in the BnaTNDH and BnaN-BC4F2populations(Fig.5).The low recombination rate observed was due mainly to the introgressed segment from Tapidor containing the low-glucosinolate alleles.The high recombination observed in this region may have been driven by the cross with NY7(high glucosinolate without the fixation of low-glucosinonate alleles)and subsequent introgression of the corresponding NY7 segment,which contains the genome segment originating in the founder B.rapa parent CDA(Fig.5).This result suggests that strong breeding selection in ancestral generations may contribute to genomic segments becoming fixed in regions accounting for target traits;that is,recurrent breeding selection through successive generations for a trait may lead to corresponding low recombination rate in the genome regions controlling the trait during the breeding process.However,it was also apparent that recombination in this region was promptly reactivated by crossing with a different genetic cluster.

4.Discussion

4.1.Common features of genome-wide meiotic crossovers and contributions of recent A-genome introgression

There are rarely more than three crossovers per chromosome per meiosis[40],an observation consistent with the known number of chiasmata.Although there was evidence of more than three recombination events in some chromosomes in some lines from the three populations,the mean number of recombination events per chromosome was still less than three,an outcome validated by random sampling of markers(Table S7).This behavior is consistent with four previously published BnaTNDH genetic maps[22,24]and the independent B.napus DH population derived from the cross of 1167×HZ396[37].

In this study,we achieved genome-wide reciprocal introgression between the two genetic clusters of semi-winter Chinese and winter European rapeseed(Fig.S1).Combining recombination information from the BnaTN populations,recombination hotspots and coldspots were identified.Recombination hotspots and coldspots provide indications of selection footprints and potential target regions for genetic improvement,and hotspots are conducive to rapid reciprocal introgression between the two genetic clusters.Thus,these hotspots and coldspots,and QTL located in these regions are expected to provide useful information for introgression breeding between the two genetic clusters.56% of conserved recombination coldspots located within centromeres,suggests that the genetic distance between the parents has only a weak effect on recombination coldspots.The detection of only seven conserved recombination hotspots and 16 coldspots in the populations,suggests that cross design influences the location of hotspots and coldspots.The finding that hotspots were more prevalent on the A than on the C genome suggests that the C genome offers higher potential for modification and breeding improvement than the A genome.

4.2.Recombination affected by genetic background and indications of crossing between the two genetic clusters

We detected differences in genome-wide θ in the BnaTNDH population compared with the two derived reciprocal introgression populations.Although only a third of the available DH lines were used to construct the backcross populations,these were selected to maximize recombination coverage in the original population,so that the crossovers sampled in these DH lines were representative of the distribution in the entire DH population(Fig.S1).Despite differences in genotyping methods,marker distributions in relation to the NY7 genome were similar(Fig.2).Although the BnaT-BC4F2population had the maximum number of genetic bins and the BnaTNDH population the fewest,the mean number of recombination events per line was similar in the two populations.This finding indicates that differences in genotyping methods did not strongly influence the assessment of θ.Although a few genome regions may still be affected by genotyping methods,the pattern of recombination appears to be retained.

The phenomenon of heterochiasmy,variation in crossovers between male and female meioses,has been detected in many eukaryotic taxa,including plants[13,41].For the latter,the reported variation may be attributable either to genetic background or to specific cross combinations,such as in a single biparental population of B.oleracea that showed higher θ arising from female meioses[13].We found no great difference in θ between BnaTNDH(3.61 cM Mb-1)and BnaT-BC4F2(3.35 cM Mb-1),but a decrease in BnaN-BC4F2(2.53 cM Mb-1).Genetic background might affect the number and distribution of recombination hotspots in the three populations,although there was no significant difference in the mean θ of hotspots(BnaTNDH:15.28 cM Mb-1;BnaT-BC4F2:15.45 cM Mb-1)(Table 2).Although the mean θ of hotspots was significantly lower in BnaN-BC4F2(11.57 cM Mb-1)than in BnaTNDH and BnaT-BC4F2,there was no significant difference in the number of recombination hotspots(Table 2).These results also suggest that the introgression of genetic diversity of European winter rapeseed in Chinese semi-winter rapeseed was relatively easy to achieve by cross breeding under the semi-winter environment,where the latter may also have gained from introgression of diploid A-genome chromosomal segments.More generally,such results also suggest that the exotic introgression of related species can increase both recombination[39]and genetic diversity.However,further analysis would be performed to test if this would be affected by the growing environment since Tapidor is a wintertype cultivar which may be affected on the cross-ability when grown in the semi-winter environment.

Fig.4.Fine mapping of flowering time,seed erucic acid content,and glucosinolate content loci with the reciprocal backcross populations.(A)Fine mapping of a flowering time locus.(B)Fine mapping of a seed erucic acid content locus.(C)Fine mapping of a seed glucosinolate content locus.FT,flowering time;ERU,erucic acid content;GLU,glucosinolate content;The black,white and gray boxes indicate NY7,Tapidor,and heterozygotus genotypes,respectively.*,P＜0.05;**,P＜0.01;ns,not significant.

Fig.5.Changes in recombination rate in three BnaTN populations with the effects of IBD inheritance on chromosome A09.(A)Recombination rate(θ)variation on chromosome A09.(B)Identity-by-descent(IBD)inheritance pattern for the NY7 pedigree.(C)IBD inheritance pattern for the Tapidor pedigree.(D)QTL for four classes of traits detected on chromosome A09.

4.3.Recombination contributed to QTL detection and its consequent effects with breeding selection

Abundant phenotypic variation was observed in the reciprocal backcross populations,including flowering time and seed quality traits.High-throughput genotyping methods were also used to genotype the populations and construct the high-density genetic linkage maps.Genome-wide association and linkage analyses have been used to map QTL for multiple traits in rapeseed,such as flowering time,erucic acid content,and glucosinolate content.However,in general the QTL intervals are typically large[20,42–44].Using the backcross populations,we succeeded in increasing the mapping resolution of QTL controlling flowering time,seed erucic acid content and glucosinolate content compared with other published populations[20,42–44].Candidate genes that may control each trait have been identified,including BnaC02.MYB28 and BnaC02.MYB34,which encode homologs of transcription factors known to regulate glucosinolate metabolism in Arabidopsis thaliana[45].This result indicates that increased recombination can occur in genome regions that control corresponding traits in backcross populations,thus improving QTL mapping resolution.The density of crossovers and genome coverage suggests that the reciprocal backcross populations are valuable resources for the fine mapping of genes controlling a wide range of important traits in rapeseed.Markers associated with candidate genes contributing to such traits are available for molecular marker-assisted breeding.Hundreds of QTL have been mapped in the BnaTNDH population along with assembled genome[7,20,21],which help with fine mapping of the QTL using the backcross populations,and these populations further provide valuable genetic resources to the

Brassica.

Genomic regions with high θ contribute to trait improvement.In contrast,strong artificial selection for a trait can also change the recombination rate of corresponding regions.Some favorable alleles in BnaTN hotspots and others that overlap with recombination hotspots.The three SNP loci in the chromosome A10 hotspot overlapped with loci reported[46]to account for variation in seed oil content.These identical regions have also been detected as selective regions in a diversity panel[8,15].It is of practical value that during selective breeding of B.napus,regions of high recombination may allow the rapid accumulation of favorable alleles that control agronomic traits.In cotton(Gossypium hirsutum L.),QTL for fiber quality traits also tended to be more common in highrecombination regions,suggesting that recombination contributes to the improvement of cotton[47].Selection can affect recombination rate[48,49].In Drosophila melanogaster,selection reduced the recombination rate[48].Strong breeding selection for a trait can lead to fixation of the associated genomic regions,resulting in corresponding low recombination rate in such regions during breeding processes.When these regions have the opportunity to be recombined into unrelated cultivars,they appear more prone to recombination.Early breeding selection on the region of A09 controlling seed glucosinolate content was detected in Tapidor[22],apparently resulting in a coldspot,whereas the same region was associated with a recombination hotspot in the NY7 pedigree parents(lacking selection for the glucosinolate trait).This region of NY7 contained the introgression of B.rapa CDA,which contributed to the increase of the genetic diversity and recombination(Fig.5).This finding is promising for gene pools and loci that have been subjected to strong selection bottlenecks,as allelic diversity may readily be broadened where increased recombination is promoted by use of germplasm that has not undergone the same selection.

We have demonstrated that a recombination map based on biparental segregating and reciprocal backcross populations provides a useful platform to identify genomic constraints on meiotic crossovers and selection strategies.The map provides information for the genomic improvement of crops,particularly in terms of the ability to accumulate and select allele combinations in multiple chromosomal regions.We have shown how high recombination regions have historically contributed to breeding selection in B.napus,and that breeding selection can also affect recombination rate.We have shown that hotspot and coldspot regions may be reactivated via cross design.The recombination variation map constructed in this study provides a reference map for gene recombination and introgression breeding between two genetic pools.Our results also provide a framework for investigating how specific recombination patterns contribute to trait improvement and be harnessed for rapeseed breeding,allelic introgression,and the accumulation of favorable allele combinations for crop breeding.

CRediT authorship contribution statement

Meng Wang:Writing–original draft,Investigation,Formal analysis,Visualization.Graham J.King:Writing–review&editing.Lei Shi:Resources.Ruiyuan Li:Resources.Yi Zhang:Formal analysis.Xiaohua Wang:Writing–review & editing.Jinling Meng:Resources.Jinxing Tu:Writing–review & editing,Funding acquisition.Jun Zou:Writing–review & editing,Supervision,Project administration,Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported by the National Key Research and Development Program of China(2021YFF1000100),the National Natural Science Foundation of China(31970564 and 32171982),and the National Key Research and Development Program of China(2016YFD0100305).We acknowledge Prof.Sihai Yang of Nanjing University for his suggestions on revising the manuscript.

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2022.07.008.