Allopolyploidization increases genetic recombination in the ancestral diploid D genome during wheat evolution

Hongshen Wn,Jun Li,Shengwei M,Fn Yng,Ling Chi,Zehou Liu,b,Qin Wng,b,Zongjun Pu,b,Wuyun Yng,b,*

a Crop Research Institute,Sichuan Academy of Agricultural Sciences,Chengdu 610066,Sichuan,China

b Key Laboratory of Wheat Biology and Genetic Improvement on Southwestern China,Ministry of Agriculture and Rural Affairs,Chengdu 610066,Sichuan,China

c The Applied Plant Genomics Lab,College of Agricultural Sciences,Nanjing Agricultural University,Nanjing 210095,Jiangsu,China

d Triticeae Research Institute,Sichuan Agricultural University,Chengdu 611130,Sichuan,China

Keywords:Synthetic wheat Hexaploidization Diploid D genome Genetic recombination Acceleration of evolution

ABSTRACT Genetic recombination produces new allelic combinations,thereby introducing variation for domestication.Allopolyploidization has increased the evolutionary potential of hexaploid common wheat by conferring the advantages of heterosis and gene redundancy,but whether a relationship exists between allopolyploidization and genetic recombination is currently unknown.To study the impact of allopolyploidization on genetic recombination in the ancestral D genome of wheat,we generated new synthetic hexaploid wheats by crossing tetraploid Triticum turgidum with multiple diploid Aegilops tauschii accessions,with subsequent chromosome doubling,to simulate the evolutionary hexaploidization process.Using the DArT-Seq approach,we determined the genotypes of two new synthetic hexaploid wheats with their parents,F2 plants in a diploid population(2x,D1D1×D2D2)and its new synthetic hexaploid wheatderived population(6x,AABBD1D1×AABBD2D2).About 11%of detected SNP loci spanning the D genome of Ae.tauschii were eliminated after allohexaploidization,and the degree of segregation distortion was increased in their hexaploid offspring from the F2 generation.Based on codominant genotypes,the mean genetic interval length and recombination frequency between pairs of adjacent and linked SNPs on D genome of the hexaploid F2 population were 2.3 fold greater than those in the diploid F2 population,and the recombination frequency of Ae.tauschii was increased by their hexaploidization with T.turgidum.In conclusion,allopolyploidization increases genetic recombination of the ancestral diploid D genome of wheat,and DNA elimination and increased segregation distortion also occur after allopolyploidization.Increased genetic recombination could have produced more new allelic combinations subject to natural or artificial selection,helping wheat to spread rapidly to become a major global crop and thereby accelerating the evolution of wheat via hexaploidization.

1.Introduction

Polyploidization has occurred during the evolution of many crops.On the basis of origin and ploidy level,polyploids are classified as autopolyploids,such as potato (Solanum tuberosum,2n=4x=48),and allopolyploids,including oilseed rape (Brassica napus,2n=4x=38,genome formula AACC),common wheat(Triticum aestivum,2n=6x=42,AABBDD),and synthetic hexaploid or octoploid Triticale (2n=6x=42,AABBRR,or 2n=8x=56,AABBDDRR) [1,2].Polyploids benefit from the advantages of heterosis (in wheat:[3,4]) and gene redundancy [5],which promote their diversity,plasticity,and adaptation after disadvantages arising from difficulties in the normal completion of meiosis (in wheat:[6,7]) and genetic instability [8–10] are overcome.Heterosis causes polyploids to be more vigorous than their diploid progenitors,whereas gene redundancy shields polyploids from the deleterious effect of mutations,thereby accelerating their evolution [5,11].

Aegilops tauschii (AT),the source of the D genome of hexaploid common wheat,is a widely distributed,genetically diverse species[12,13].It serves as a gene reservoir for modern wheat adaptation[4,14,15],as its superior resistance to disease [16–19] and pests[20–22],tolerance to environmental stresses[23–26],and improving yield-related traits [27–30]should theoretically be manifested in hexaploid wheat via heterosis.Approximately 9000 years ago,an accidental hybridization added its genome to emmer wheat,resulting in the generation of a free-threshing form:hexaploid common wheat (bread wheat,T.aestivum),which now accounts for approximately 95%of global wheat production,with tetraploid durum wheat(T.turgidum ssp.durum) representing the remaining 5% [31].Four morphological variants of Ae.tauschii have been found in nature.Three of the variants:typica,anathera,and meyeri,are grouped into Ae.tauschii ssp.tauschii,while the fourth is assigned to Ae.tauschii ssp.strangulata [32].However,because the D genome of early bread wheat originated from only a small number of wild Ae.tauschii ssp.strangulata plants [13,33,34],the genetic basis and diversity of modern common wheat are narrow.To broaden the genetic diversity of the D genome of bread wheat,scientists have created synthetic hexaploid wheats (SHWs) by crossing tetraploid wheat with Ae.tauschii and subsequently doubling chromosomes[35].An increasing number of elite commercial wheat cultivars derived from SHWs are being released [3,36–39].Artificial hexaploidization may thus increase the genetic variation and speed the adaptive evolution of modern wheat [3] and is accordingly of increasing interest [38–41].

Genetic recombination produces novel allelic combinations,thereby generating new variation for evolution.Genetic advances in breeding programs depend on the selection of new recombinant plants from inter-cultivar or interspecific crosses.Boosting genetic recombination should thus accelerate the combination of valuable traits from different parents in new elite cultivars [42].In plants,recombination frequency (RF) is often determined by genetic factors,such as zep1 in rice [43],zyp1 in barley [44],Ph1 in wheat[45],a related quantitative trait locus (QTL) in wheat [46],and male meiosis in wheat[47]and rye[48].Recombination frequency is also affected by ploidy level [49–51] and external stresses,such as chemicals,nutrient deficiency,X-rays,mutagenic agents,and temperature,which can change plant/cell physiological conditions[52,53].In Brassica,the genetic recombination frequency in the diploid A genome is elevated in allotriploid and allotetraploid hybrids with a C genome [54].In Arabidopsis,polyploidization increases meiotic recombination frequency [55].Allopolyploidization,a focus of interest among plant evolutionary geneticists in recent years,is a driving force in plant evolution[56,57].In wheat,hexaploids have the advantages of heterosis and gene redundancy derived from both emmer wheat and the wild grass Ae.tauschii.Studies of newly formed allopolyploids [1,2,58,59] have revealed dynamic and stochastic changes in genomic organization and gene expression,including sequence elimination,interchromosomal exchange,and cytosine methylation in wheat[60–63],gene repression[4],novel activation,genetic dominance[64,65],subfunctionalization[66],and transposon activation in wheat[67].In contrast,however,changes in genetic recombination,such as homologous recombination during polyploidization,which might strongly influence plant genome variability [68],have received little attention in investigations of the evolution of wheat,an excellent model plant for studying polyploidization [69].

In this study,to simulate the evolutionary hexaploidization process,we generated new synthetic hexaploid wheats by crossing tetraploid T.turgidum with different diploid Ae.tauschii,with subsequent chromosome doubling.And two F2segregating populations with different ploidy levels were developed using them to investigate RF and the presence of segregation distortion.DNA elimination of D genome after allohexaploidization were also examined using genotyping by sequencing (GBS).The aim of this study is identifying the effects of hexaploidization on ancestral D genomes in wheat,including DNA elimination,segregation distortion and RF,to find some important clues about the force of hexaploidization allowing wheat to overcome the limitations of its genetic narrow D genome and accelerate its adaptive evolution into a major food crop during wheat evolution.

2.Materials and methods

2.1.Plant materials

Two Ae.tauschii accessions,SQ665 and SQ783 (Ae.tauschii ssp.tauschii var.typica),and a tetraploid T.turgidum cultivar,Langdon(T.turgidum ssp.durum),were used to generate two new SHWs:Langdon/SQ665 (LS665) and Langdon/SQ783 (LS783).Seeds of the two Ae.tauschii accessions were provided by Dr.Mujeeb-Kazi of the International Wheat and Maize Improvement Center (CIMMYT),Mexico,in 1995.The seeds of Langdon were provided by the Triticeae Research Institute of Sichuan Agricultural University.Langdon was used as a female parent in crosses with the two Ae.tauschii accessions and subsequent chromosome doubling.As determined by fluorescence in situ hybridization with two repetitive DNA sequences,Oligo-pSc119.2 labeled with Alexa Fluor 488-5-dUTP and Oligo-pTa535 labeled with Texas Red-5-dCTP [70],the majority of offspring from the first selfing generation (S1) of LS665 and LS783 contained the euploid chromosome set(2n=42) karyotyped by Zhu [71].We selected karyotyped S1 offspring containing 42 chromosomes covering the entire A,B,and D genomes and created a hexaploid F2mapping population of synthetic hexaploid wheat LS665 × LS783 (SHWP,population size:182),to evaluate the RF on homologous chromosomes of D genomes in the hexaploid genetic background.The RFs of F2populations of Ae.tauschii SQ665 × SQ783 (ATP,population size:123)were calculated as a control to investigate hexaploidization changes in RF in the diploid D genomes.

2.2.SNP genotyping

The two F2populations and their parents were planted in the field using the single-seed precision method.Plant tissue (50 mg)was collected from 2-week-old seedlings and subjected to DNA extraction using a NuClean Plant Genomic DNA kit(CWBio,Beijing,China).The extracted DNAs were quantified on a Qubit 4 fluorometer (Life Technologies,Singapore),and their concentrations were adjusted to levels suitable for genotyping.

GBS analysis of the two populations together with their parents was performed using DArT-Seq technology,which combines diversity arrays technology (DArT) and second-generation sequencing[72],to detect potential SNPs by comparing generated sequences or reads with the wheat genome preassembly and wheat,rye,and Triticale consensus map v4 by Diversity Arrays Technology P/L(Bruce,ACT,Australia)in 2016.This genotyping technology relies on complexity reduction using restriction enzyme digestion (Pst I in this study) to enrich genomic representations with single-copy sequences followed by second-generation sequencing on a HiSeq instrument.Approximately 100 μL of each 50 ng μL-1DNA sample was sent to Diversity Arrays Technology P/L for GBS analyses.Only SNP data were used for linkage map construction and RF calculations,as RFs calculated using codominant genetic markers are more accurate than those based on dominant genetic markers[73,74],and 28,331 SNPs were obtained with GBS data.A set of 9043 of the 28,331 SNPs with sequences of the SNP tags were aligned to the D genome of the CS RefSeq v1.0 assembly [75] in 2018.To select SNPs that were specific to the diploid D genomes of Ae.tauschii SQ665 and SQ783,the durum wheat Langdon,which was the AB genome donor for both LS665 and LS783,was also genotyped and used to screen out any of the 28,331 SNPs that were also detected in Langdon.We also re-aligned their physical positions of these SNPs to the sequence assembly of Ae.tauschii AL8/78[15] and the other four available Ae.tauschii sequence assemblies[76] in WheatOmics v1.0 (http://202.194.139.32/blast/blast.html).When using basic local alignment search tool(BLAST),the filtering thresholds for E-value and query coverage were 0.00001 and 40%,and BLAST hits with both minimum E-value and maximum query coverage were considered to be the most likely physical position for the SNP in the reference D genome.

2.3.Genetic maps and statistical analyses

After filtering of detected SNPs,774 SNP markers (Table S1)remained for linkage map construction and RF calculations in the D genome of the two populations.The criteria used for filtering SNPs were as follows:(1) the remaining SNP markers used for D genome map construction could not be detected (were displayed as missing data)on the AB genome of Langdon;(2)each employed SNP markers was also polymorphic between the parents of both diploid and hexaploid F2populations;(3)genotypes of all involved parents were homozygous at the employed SNP sites,and the diploid parents SQ665 and SQ783 had the same genotypes as their SHW offspring LS665 and LS783,respectively;and (4) the SNP markers were able to detect both homozygous and heterozygous genotypes.

Genetic linkage map construction and maximum-likelihood calculation of the RF between two adjacent SNP markers was performed with QTL IciMapping 3.2 [74,77,78].To compare the RF between the two genetic populations,the RF between each pair of adjacent,linked SNP loci on the physical map was calculated using the‘‘a(chǎn)lgorithm by input”function of QTL IciMapping.The frequency of different parental genotypes or genes and statistics of genotypes were calculated using Microsoft (MS) Excel for Mac OS(Microsoft Corp.,Redmond,WA,USA).Chi-square and t-tests,as well as the drawing of statistical and scatter plots,were performed with MS Excel statistics and drawing functions.The stem-and-leaf plot,linear regression and analysis of variance for the regression model were executed with SPSS statistics for Mac OS (IBM Corp.,Armonk,NY,USA).

3.Results

3.1.SNP distribution and genetic map construction

DArT-Seq technology was used to genotype the D genomes of two genetic populations with different ploidy levels.A total of 9043 SNPs were obtained via GBS based on DArT-Seq technology and subsequent alignment of SNP tags (sequences from GBS) to the D genome of the CS RefSeq v1.0 assembly,and 8330 of the 9043 SNPs could also be mapped to the Ae.tauschii AL8/78 sequence assembly.The number of SNPs found on D-genome chromosomes ranged from 795 on chromosome 4D to 1850 on chromosome 2D (Table 1).The number of detected SNPs increased consistently and linearly with physical distance from the centromere of each chromosome,with the vast majority of SNPs located in distal,centromere-free regions (Fig.1).

A total of 774 polymorphic SNP markers(Table S1)in the D genome were finally selected for map construction and RF calculations of the two F2populations,and the proportion of SNP polymorphisms between the parents of the two populations was 8.6% of the 9043 SNPs(Table 1).The number of polymorphic SNP markers ranged from 71 on chromosome 4D to 143 on chromosome 2D(Table 1),and their distributions showed a linear decrease toward the centromeres(Fig.1).Almost no polymorphic SNP markers were located near the centromere;most were clustered in genomic regions far from pericentromeric regions,such as regions located 140 Mb to 160 Mb and–220 Mb to–300 Mb from the centromere of chromosome 1D (Fig.2).These SNP-rich regions spanned approximately 1520 Mb,with a mean density of ＞2 SNPs per 10 Mb (Fig.2).

The genetic maps were constructed according to SNP physical positions of the Ae.tauschii AL8/78 sequence assembly,considering the differing physical lengths among Ae.tauschii accessions,and the ‘‘nearest neighbor and two-opt” (nnTwoOpt) method [79]was also used for genetic map construction as self-organization(Table 1).For the diploid population,the genetic map constructed using the nnTwoOpt method was 3226.7 cM long,and mapped lengths of individual chromosomes ranged from 251.4 cM for 4D to 548.8 cM for 2D,with one linkage gap on chromosome 3D(Table 1).The genetic map constructed using the physical map of Ae.tauschii AL8/78 was 110 cM (3.3%) longer than that by the nnTwoOpt method,without linkage gaps on the chromosomes(Table 1).For the hexaploid population,the total length of the genetic map constructed using the nnTwoOpt method was 6978.7 cM,with chromosome lengths ranging from 515.0 cM for 4D to 1260.7 cM for 7D,while the genetic length increased about 210 cM (3.0%) using SNP physical alignment (Table 1).No linkage gap was found in either hexaploid genetic map.The difference in total genetic map length between the two methods was less than 5.0% in both diploid and hexaploid populations.

3.2.DNA elimination and segregation distortion

For Ae.tauschii accession SQ665,4588 SNPs were detected in its genome,without any genotypic data detected in Langdon(Table S2).The Ae.tauschii AL8/78 sequence assembly [15],the D genome of the CS RefSeq v1.0 assembly [75] and four available Ae.tauschii sequence assemblies of AY17,AY61,T093 and XJ02[76] to align the chromosome positions of 4588 SNPs.After hexaploidization with Langdon,483 (10.5%) of the SNP loci of SQ665 were eliminated in the hexaploid LS665(Table 2).The elimination rates on chromosomes ranged from 8.0%on 4D to 11.7%on 3D(Table 2).Among the 4566 SNPs detected on SQ783,525(11.5%)were eliminated in its hexaploid LS783(Table 2),and the elimination rates on chromosomes ranged from 8.23%on 4D to 13.52%on6D (Table 2).However,among the detected SNPs on SQ665 and SQ783,356 and 331 loci could not be anchored to any chromosomes of Ae.tauschii AL8/78,AY17,AY61,T093,XJ02,and the D genome of CS,and respectively 9.8%and 11.8%of them were eliminated after hexaploidization(Table 2).At the detected SNP loci on Ae.tauschii,only 114 of 3705 overlapping SNP loci in both SQ665 and SQ783 were eliminated in both SHW LS665 and LS783(Table S2).More SNP loci were eliminated in SQ783 than SQ665 after their hexaploidization with Langdon.Among the eliminated 894 SNP loci in their derived F2populations,including 114 overlapping eliminated SNPs,the number of loci at which frequencies of the detectable genotype,including genotypes AA (homozygous male type),AB (heterozygous type),and BB (homozygous female type),ranged from 90% to 100%,was approximately 480 in the Ae.tauschii population,much greater than the number (about 80)of eliminated SNP loci on their hexaploid population (Fig.3).At the eliminated SNP loci,the mean frequency of detectable genotypes in the Ae.tauschii population was significantly larger than that in SHW population (Fig.3).

Table 1 Genetic map lengths of chromosomes of the D genome aligned with the physical map of AL8/78 and automatically organized by the nnTwoOpt method.

Fig.1.SNP distribution in the D genome based on a statistical window size of 10 Mb.The centromere of each chromosome is symbolized by a dotted line.The x axis indicates the starting position of each 10-Mb window in the D genome,and the y axis indicates the number of SNP markers in each 10-Mb region.

Fig.2.Distribution of polymorphic SNP markers in D genomes.The physical position of the centromere on each chromosome is set to 0 Mb.

With the genotype data of 774 SNPs in populations,segregation distortion loci(SDL)were also investigated in both the ATP and the SHWP.Respectively 125 and 396 of 774 SNP loci in the D genome showed significant segregation distortion at the P=0.01 level in ATP and SHWP (Table 3).The number of SDLs ranged from 4 on 5D to 56 on 7D in ATP,and most of the SDLs on chromosome 1D,2D,3D and 6D were biased towards the male parent SQ783,whereas SDLs on chromosome 4D and 7D were mostly biased towards the female parent SQ665(Table 3).On SHWP,the number of SDLs ranged from 46 on 3D to 75 on 2D.The majority of SDLs on chromosomes 2D,4D,and 6D were biased towards the male parent LS783,and those on chromosomes 1D,3D,5D,and 7D were mostly biased towards the female parent LS665 (Table 3).

The P-values of chi-square tests of observed against theoretical segregation ratios at most of the SNP loci in the SHWP were much higher than those in the Ae.tauschii population (Fig.4A),and the level of segregation distortion across the whole diploid D genome on F2population was greatly increased after hexaploidization.The parental gene frequency was also changed after hexaploidization,and at most of the SNP loci of chromosome 1D and 3D,the gene frequency of the male parent originating from SQ783 [p (A)]became less than that of the female parent originating from SQ665[p(B)]in the SHW population,whereas the gene frequency of the male parent was greater than that of the female parent in the diploid population(Fig.4).On chromosome 4D,the change in segregation distortion level was much larger than that on other chromosomes,as the mean [log0.05(P-value) -1] of all SNP loci on SHWP was greater than that on other chromosomes (Table S1;Fig.4A).Otherwise,of the 71 SNP loci on chromosome 4D,61 SNP loci with significant segregation distortion had much higher gene frequencies from LS783 alleles than those from LS665 in SHWP (Table 3;Fig.4B).

3.3.The level of genetic recombination in diploid D genomes after hexaploidization

The 774 polymorphic SNP markers(Table S1)used for map construction and RF calculations spanned a physical length of 3992.9 Mb in the Ae.tauschii genome and the chromosome lengthsranged from 493.8 Mb for chromosome 6D to 648.7 Mb for chromosome 2D (Table 4).The genetic map constructed using the diploid population was 3334.9 cM long,and mapped lengths of individual chromosomes ranged from 263.2 cM for 4D to 593.3 cM for 2D (Table 1;Fig.5).The total length of the genetic map constructed using the hexaploid population was 7185.4 cM,a value approximately two fold greater than that of the diploid genetic map (Table 1;Fig.5).This situation was also true of the individual chromosome maps,with the mapped length of each chromosome in the hexaploid ranging from 519.8 cM for 4D to 1279.4 cM for 7D (Table 1;Fig.5).

Table 2 Elimination of SNP alleles in D genomes induced by hexaploidization in synthetic wheats of LS665 and LS783.

Fig.3.Distribution of detectable-genotype frequency at eliminated SNP loci in Ae.tauschii and synthetic hexaploid wheat populations.The detectable-genotype frequency at one SNP locus is calculated as:1 -number of missing genotypes/population size.

The mean genetic distance between pairs of adjacent and linked SNP loci on each chromosome of the diploid Ae.tauschii population ranged from 3.76 cM on 4D to 5.39 cM on 1D,whereas these distances varied from 7.83 cM on 4D to 13.67 cM on 1D in the SHW-derived population (Table 4).The mean genetic distance between adjacent SNP loci on each chromosome in the hexaploid genetic background was significantly greater than that in the diploid;the–log (P) generated by a two-tailed paired-sample ttest of the difference ranged from 12.40 to 35.55,and the mean genetic distance ratio of SHWP to ATP varied between 1.98 and 2.54(Table 4).Across the entire ancestral diploid D genome,which comprised 767 intervals,the mean genetic length was increased by 2.33 fold,from 4.35 in ATP to 10.13 in SHWP(Table 4).The value of ΔSHW-ATbetween hexaploid and diploid populations ranged from 4.07 cM on chromosome 4D to 8.28 cM on chromosome 1D(Table 4).The mean RF between adjacent SNP loci on each chromosome ranged from 3.74% on 4D to 5.34% on 1D in ATP and from 7.70% to 13.17% on these respective chromosomes in SHWP(Table 4).The mean RF of each chromosome in SHWP was significantly higher than that in ATP,with–log(P)ranging from 12.69 to 36.74 and their ratios varying from 1.96 to 2.47 (Table 4).The mean ΔSHW-ATof the RF of each chromosome between different ploidy-level populations ranged from 3.96% on 4D to 7.83% on 1D (Table 4).

Along the entire D genome,the difference in RF between synthetic hexaploid wheat and Ae.tauschii (ΔRFSHW-AT) varied from–5.0%to 17.0%(Fig.6;Table S3),with a mean of 5.6%and a RFSHWP/RFATPratio of 2.3 (Table 4).As revealed by a stem-and-leaf plot of the distribution of ΔRFSHW-ATvalues,seven outliers were also observed in some genomic regions.In the SNP0789–SNP0109 genomic region corresponding to a physical position of 616.8 Mb to 618.1 Mb on chromosome 2DL,the RF of the diploid increased by 18.0% (5-fold) — from 4.5% to 22.5% — after hexaploidization(Fig.6;Table S3).The RF of the region from SNP1036(physical position:369.3 Mb) to SNP1351 (371.1 Mb) on 3D was increased by 18.2% (about 10-fold) in the hexaploid (Fig.6;Table S3).RFs of the intervals SNP0962–SNP0963 from 561.6 Mb to 562.2 Mb onchromosome 5D (Fig.6;Table S3) and SNP1044–SNP0252 (Fig.6;Table S3) from 5.3 Mb to 7.2 Mb on chromosome 7D increased by respectively 18.2% (6.5-fold) and 18.6% (3-fold) after hexaploidization (Fig.6;Table S3).In contrast,three genomic regions were characterized by RFs that had decreased by ＞6.0%after hexaploidization.In the interval corresponding to physical positions between 45.7 Mb (SNP1418) and 58.3 Mb (SNP0527) on chromosome 5DS,the RF decreased by 10.7% after hexaploidization,and the intervals SNP0177–SNP0544 from 17.6 Mb to 20.8 Mb on 4D and SNP1234–SNP0942 from 352.3 Mb to 364.7 Mb on 5D had RFs that were respectively 6.0% and 6.5%lower after hexaploidization (Fig.6;Table S3).

Table 3 Segregation distortion loci (SDL) in SQ665 × SQ783 AT F2 population and LS665 × LS783 SHW F2 population.

Table 4 Genetic interval length and recombination frequency (RF) between pairs of adjacent,linked SNP loci in the D genome in ATP and SHWP.

Fig.4.The segregation distortion(A)and parental gene bias(B)of SNP sites in the ancestral D genome in AT and their SHW populations.P-value was calculated by chi-square test with theoretical genotype frequency of 1:2:1(AA:AB:BB).IF the[log0.05(P-value)-1]＞0 or 0.54 at a SNP site,significant segregation distortion was inferred at this locus at P=0.05 or 0.01 level,respectively.p(A)means gene frequency of the male parent SQ783 at one SNP site;p(B)means gene frequency of the female parent SQ665.If(p(A)/p(B)– 1) ＞0,the gene frequency of SQ783 was greater than that of SQ665 in the F2 population,and the genotype at this locus was biased toward SQ783.

Overall,the mean RF between adjacent and linked SNP loci in the hexaploid was significantly higher,by approximately 2.3 fold,than in the diploid,and the mean RF of the diploid increased from 4.31% in ATP to 9.87% in SHWP after hexaploidization (Table 4).A significant linear relationship was observed between RF of ATP (x)and RF of SHWP (y),and the regression equation of y=0.9118x+0.0594 was significant(R2=34.5%,P less than 0.0001)(Fig.7A).However,there was no significant linear relationship between ΔRFSHW-ATand the physical distance between linked and adjacent SNP loci (Fig.7B),indicating that the RF change of the ancestral diploid D genome after hexaploidization was independent of the SNP density of the D genome.

4.Discussion

4.1.Genetic map construction based on SNP physical positions

Fig.5.Collinearity between genetic and physical positions in the two diploids and their SHW-derived populations.The x axis indicates the physical position of each SNP in the D genome and the y axis indicates its genetic position.The centromere of each chromosome is indicated by a dotted line at its physical position.

Fig.6.Change in RF between diploid Ae.tauschii accessions and their SHW-derived F2 populations.The x axis indicates the physical position of the right boundary of the interval between two adjacent,linked SNP loci.The y axis is the difference in RF between the hexaploid F2 population of SHW and the diploid F2 population of AT(ΔRFSHW-AT).

The complexity-reduction method for GBS (DArT-Seq) used in this study was based on restriction enzyme digestions of Pst I(cleaving TGCAG at the front of SNP tags),which is a methylation-sensitive restriction enzyme [72].When Pst I is used for complexity reduction,the ‘‘methylation filtration” effect will reduce the genomic representation of hypermethylated regions[72,80],such as the centromeric region[81],where there are fewer cleavage sites of Pst I.For this reason,the SNP tags covered many fewer sequences in centromeric than in telomeric regions (Figs.1,2).But the methylation-sensitive enzymes more effectively enriched for SNPs located in genic regions than methylationinsensitive enzymes such as Mse I[80],given that more SNPs were detected in in distal,centromere-free regions.

One linkage gap was found on chromosome 3D,on the chromosome genetic maps generated by the nnTwoOpt method(Table 1).Considering that different Ae.tauschii accessions often differ in genome size [76],we used only one of their reference genomes for alignment.Taking into account that only slight intervallength differences were observed between the nnTwoOptorganized genetic map and the AL8/78-physically aligned map(Table 1),we used only the AL8/78 physical maps [15] to align SNPs to the D genomes of SQ665 and SQ783.Using this physical map as an alignment reference,comparative changes in genetic distances and RFs in the diploid D genome after hexaploidization could thus be directly and conveniently investigated between populations with different ploidy levels.

SDLs have been observed in many Ae.tauschii and SHW populations [30,82,83].In our study,the number of SDLs and the level of segregation distortion in D genomes increased after hexaploidization,possibly as a result of genomic reorganizational events,such as sequence elimination [60,61] and interchromosomal exchange[62],after allopolyploidization.Even though different levels of segregation distortion were detected between ancestral D genomes and their allohexaploids,we suggest that this phenomenon had only subtle effects on our genetic map construction and RF calculations.According to previous authors [84,85],segregation distortion has little or no influence on linkage map construction,and estimation of RF between codominant markers is less affected by selection than that between dominant markers [73].

Fig.7.Scatter plots with linear regression relationships between (A) RF of ATP and RF of SHWP (B) Interval physical length of ATP and ΔRFSHW-AT.RF,recombination frequency;ATP,F2 populations of Ae.tauschii SQ665 × SQ783;SHWP,F2 population of synthetic hexaploid wheat LS665 × LS783 SHWP;ΔRFSHW-AT,the difference in RF between SHWP and ATP.

DNA elimination of the diploid D genome also occurred in their hexaploids,in agreement with previous studies [60,61],indicating rapid genomic changes in newly formed hexaploid wheat.For the two newly synthesized hexaploid wheats LS665 and LS783,among their eliminated SNP loci,only 114 were eliminated in both SHWs,suggesting that DNA elimination might occur randomly at the majority of SNP loci in the diploid D genome after hexaploidization.So at the eliminated SNP loci in LS665 and LS783,genotypes of the alleles present were still detected in the LS665 × LS783 F2population,and the mean frequency of the detectable genotypes was about 50% (Fig.3).This frequency was a result of only 114 overlapping SNPs of the total 894 eliminated SNPs in the two parents,and the frequencies of the detectable genotypes of the 114 overlapping eliminated SNPs were less than 30% (Fig.3).As about 20% of the genotypes were also detected in the 114 overlapping eliminated SNPs of SHWP on average,we inferred that not all of the F2plants were derived from only one plant of the first selfing generation from Langdon/SQ665 or Langdon/SQ783 containing 42 chromosomes covering the entire A,B,and D genomes.Different S1 plants from a single hybrid cross made for hexaploidization might lose DNA sequences randomly.

Some eliminated SNPs were a result of putative differences in DNA methylation levels in genomic regions of Ae.tauschii and their hexaploids.Despite no significant difference in mean DNA methylation level between Ae.tauschii and its hexaploid [63],methylation levels in specific genome regions of Ae.tauschii increased after its hexaploidization [86].Such an increase might lead to different cleavage efficiencies of Pst I in the same DNA sequence[72],and consequently rare sequencing reads by DArT-Seq covered these genome regions in the hexaploids.However,Shaked et al.[9] found that one genome can eliminate up to 14% of its loci in a single generation,as the one of the major and immediate responses of the wheat genome to allopolyploidy,in comparison with only 4.4% loci maintaining methylation pattern alteration in allopolyploids.We propose that the majority of the observed eliminated SNPs were due to genetic rather than epigenetic alterations.

Considering the disadvantage of the Pst I-dependent complexity-reduction method for GBS,further studies using whole-genome resequencing would describe more accurately the pattern of DNA elimination.

4.2.Increased genetic recombination in the ancestral D genome after hexaploidization

To examine changes at the chromosome level,we compared the mean RF between adjacent,linked SNP loci in the diploid population with that in the SHW population.An increase in RF was observed in the ancestral diploid D genome after hexaploidization.In agreement with previous reports of a more strongly altered D genome than AB genome [61],the greater change (genomic instability) of the D genome may be attributed to the increase in RF in the ancestral diploid D genome after allohexaploidization.DNA elimination was found in the D genome of both Ae tauschii SQ665 and SQ783 after hexaploidization,suggesting that DNA elimination could be attributed to genomic instability of the ancestral diploid D genome after hexaploidization.

Most previous studies have focused on changes in RF accompanying a shift from the diploid to the tetraploid state.Leflon et al.[54] reported that the total genetic length of the A7 linkage group increased from 52 cM in the diploid Brassica rapa to 96 cM in the tetraploid B.napus following the allopolyploidization event that gave rise to the latter species.The meiotic RF increased from 15.4% in the diploid Arabidopsis thaliana to 24.1% in the allotetraploid A.suecica [55].These reports suggest that the RF of the diploid genome will increase by less than two fold [55] following polyploidization from diploid to tetraploid.In the present study,the RF of the diploid genome increased by more than twofold(Table 4) when the ploidy level was raised from diploid to hexaploid.These results indicate that the extent of the RF increase of the ancestral genome depends on the polyploidization level,as the rate of RF increase from a diploid to a hexaploid state was much greater than that accompanying a diploid-to-tetraploid transformation.This dependence on polyploidization level,however,might be applicable to euploids but not to aneuploids.In B.rapa,for example,the total genetic length of the A7 linkage group of the allotriploid was fourfold higher than that of the diploid and tetraploid [54].In two pentaploid wheat F2populations,the RF of the AB genome of the pentaploids was fourfold higher than that in hexaploid wheat (unpublished data),much higher than that of the diploid in this study.Two possible reasons for these observations can be suggested.First,aneuploidy causes greater genomic instability than polyploidy(euploidy)in organisms[2,87],and aneuploidy itself can be responsible for the propagation of chromosomal instability [88,89].Second,chromosomes that remain as univalents in the aneuploid can cause a compensatory increase in crossover frequency among unaffected bivalents [54,90-92].

Although the RF of the diploid D genome was increased after hexaploidization,the responsible genetic mechanisms have not yet been elucidated.However,reported QTL affecting crossover(CO) frequency have been detected in hexaploid wheat,and most were distributed on the AB genome.Using 13 recombinant inbred mapping populations,Gardiner et al.[93]detected five QTL for CO frequency on chromosomes 2A,2B,4B,5A,and 6A of the AB genome of common wheat,and Jordan et al.[46] detected 40 QTL for total CO frequency,also mostly on the A and B genomes,by nested association mapping.These results indicate that the genetic factors determining CO frequency are present in the ancestral AB genome and are also responsible for the increase in the RF of the diploid DD genome after hexaploidization.The maximum phenotypic effect of the QTL reported by Gardiner et al.[93] was an increase in CO frequency of less than 15%,however,and the 40 QTL detected across the entire genome by Jordan et al.[46]increased the overall mean number of COs by only 7.0%.The effect size of these QTL was much lower than the extent of the RF increase of the diploid genome caused by hexaploidization in our study.This observation suggests that the RF increase of the diploid D genome in our study was due mainly to hexaploidization,whereas the contribution of the genetic factors on the AB genome were minor even though the AB subgenome predominates over the D subgenome [64,94,95].

Ghani [96] compared DNA methylation levels among B.rapa(AA),B.nigra (BB),and their allotetraploids (AABB) using a methylation-sensitive amplification polymorphism approach.They found that DNA methylation levels decreased with increasing ploidy levels [96],while recombination was promoted in hypomethylated chromosome arms [97].In wheat,no significant difference in genome-wide DNA methylation levels was observed between Ae.tauschii and the D genome of a resynthesized hexaploid wheat [63].It is thus hard to infer that the RF increase was a consequence of DNA methylation changes after hexaploidization.

4.3.Polyploidization increases the variation and accelerates the adaptive evolution of bread wheat

Although the AB subgenome predominates over the D subgenome,hexaploid wheat has a heterotic advantage over both tetraploid wheat and Ae.tauschii,as polyploidization increases heterobeltiosis for adaptive traits,such as salt tolerance,in newly synthesized allohexaploids [98].Firstly,Ae.tauschii genes can be expressed normally in a hexaploid genetic background,given that many reported QTL/genes have been mapped to the D genome of synthetic hexaploid wheat [16–22,25,26,29,30,99].And mRNA and small-RNA transcriptome analyses of nascent hexaploid wheat have revealed the generation of heterosis in common wheat[4].As allopolyploidization has accelerated the evolution of wheat by this main way,our findings suggest that the increased genetic recombination of the ancestral diploid D genome caused by hexaploidization can be regarded as a new advantage of allopolyploids,accelerating their evolution.

The added D genome conferred improved adaptability to changing environments and accelerated the spread of the hexaploid,however,the D genome of the earliest bread wheat,was derived from only a small fraction of the Ae.tauschii population that could not have fortuitously harbored all of the superior traits of different Ae.tauschii variants.We hypothesized that some other mechanism(s) were responsible for the more rapid spread of hexaploid common wheat than of tetraploid wheat.Our study has revealed that hexaploidization increased genetic recombination in the ancestral diploid D genome in allohexaploid wheat.The RF of the entire D genome of SHW was 2.3 fold higher than that of the diploid;this would increase genetic variation in intercrosses among plants of early hexaploid wheat,as an elevated number of recombination events has the potential to accelerate the development of new cultivars by(1)allowing the quick assembly of novel beneficial multiallelic complexes and (2) breaking the linkage among unfavorable genes and fixing desirable genes in fewer generations [42].This process was more efficient than that in a diploid genetic background,as more recombination events occurred in the hexaploid genetic background,with a higher possibility to incorporate more phenotypic variation.

5.Conclusions

We simulated the evolutionary hexaploidization process of wheat by crossing tetraploid T.turgidum with Ae.tauschii with subsequent chromosome doubling.DNA elimination and increased segregation distortion occurred in the ancestral diploid D genome after their hexaploidization.Comparison of the recombination frequency in diploid populations with that in their SHW populations showed that the genetic recombination in the ancestral diploid D genome was increased by 2.3 fold after hexaploidization.We propose that the increased number of recombination events could generate more new types of gene combinations subject to natural or artificial selection,resulting in accelerated adaptive evolution of wheat.

CRediT authorship contribution statement

Hongshen Wan:Writing– original draft,Formal analysis,Data curation,Funding acquisition.Jun Li:Resources,Methodology,Funding acquisition.Shengwei Ma:Data curation.Fan Yang:Investigation.Liang Chai:Visualization.Zehou Liu:Software.Qin Wang:Project administration.Zongjun Pu:Funding acquisition.Wuyun Yang:Conceptualization,Supervision,Funding acquisition,Writing– review &editing.

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 study was partially supported by the Accurate Identification Project of Crop Germplasm from Sichuan Provincial Finance Department,the National Natural Science Foundation of China(31661143007),the National Key Research and Development Program of China (2017YFD0100900),the Science and Technology Department of Sichuan Province (2017JY0077,2021YFYZ0020,and 2021YFYZ0002),and the Sichuan Provincial Finance Department (2019QYXK034).We are very grateful to Prof.Baorong Lu and Zhiyong Liu,respectively from Fudan University and Chinese Academy of Sciences for the critical review of this manuscript.

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2021.09.002.