Reproductive and genetic roles of the maternal progenitor in the origin of common wheat (Triticum aestivum L.)

Abstract Common wheat (Triticum aestivum L., AABBDD genome) is thought to have emerged through natural hybridization between Triticum turgidum L. (AABB genome) and Aegilops tauschii Coss. (DD genome). Hybridization barriers and doubling of the trihaploid F1 hybrids’ genome (ABD) via unreduced gamete fusion had key roles in the process. However, how T. turgidum, the maternal progenitor, was involved in these mechanisms remains unknown. An artificial cross‐experiment using 46 cultivated and 31 wild T. turgidum accessions and a single Ae. tauschii tester with a very short genetic distance to the common wheat D genome was conducted. Cytological and quantitative trait locus analyses of F1 hybrid genome doubling were performed. The crossability and ability to cause hybrid inviability did not greatly differ between the cultivars and wild accessions. The ability to cause hybrid genome doubling was higher in the cultivars. Three novel T. turgidum loci for hybrid genome doubling, which influenced unreduced gamete production in F1 hybrids, were identified. Cultivated T. turgidum might have increased the probability of the emergence of common wheat through its enhanced ability to cause genome doubling in F1 hybrids with Ae. tauschii. The ability enhancement might have involved alterations at a relatively small number of loci.


MATSUOKA And MORI
The T. turgidum cultivars are classified into two types based on grain threshability: the hulled type, including emmer wheat [T. turgidum subsp. dicoccum (Schrank ex Schübl.) Thell.] and the free-threshing type, including durum wheat [T. turgidum subsp. durum (Desf.) Huns.]. Emmer wheat is prototypic. Lines of genetic and archeological evidence suggest that the female progenitor of common wheat likely was a free-threshing cultivar (Matsuoka, 2011). The location in which the original hybridization took place is unknown; however, the southern coastal region of the Caspian Sea is a good candidate based on genetic evidence (Matsuoka & Takumi, 2017;Tsunewaki, 1966;Wang et al., 2013).
The evolution of common wheat represents a typical case of plant allopolyploid evolution in which hybridization barriers between the parental species and unreduced gamete production in their F 1 hybrids are highly influential (Ramsey & Schemske, 1998).
Pre-and postpollination mechanisms can prevent allopolyploid speciation from occurring by limiting the formation and viability of interspecific F 1 hybrids, whereas the fusion of unreduced gametes results in the formation of nascent allopolyploids through doubling of the F 1 hybrid genome. In crosses of T. turgidum with Ae. tauschii, various hybridization barriers, such as reduced crossability between the parental species and lethality/weakness in the F 1 hybrids, are observed (Matsuoka & Takumi, 2017;Mizuno et al., , 2011Nishikawa, 1953). The trihaploid F 1 hybrids (ABD genome) undergo genome doubling through unreduced gamete fusion and set allohexaploid F 2 seeds (AABBDD; Kihara, 1946). The genotypes of parental T. turgidum and Ae. tauschii accessions influence the frequency of genome doubling, which can be measured as the seedset rates of F 1 hybrids (Fukuda & Sakamoto, 1992;Kihara et al., 1965).
Previous studies have shed some light on the role of Ae. tauschii in the evolution of common wheat. Ae. tauschii has three intraspecific lineages that are defined on the basis of genome-wide molecular marker genotypes: two major lineages (TauL1 and TauL2) and one small lineage (TauL3). TauL2 and TauL3 are more closely related to the D genome of common wheat than TauL1 (Matsuoka et al., 2013;. Geographically, TauL1 is spread across the species range, whereas TauL2 is restricted to the western region of the species range, that is, the Transcaucasian-Middle Eastern region, and TauL3 exists only in Georgia. Interestingly, Ae. tauschii is polymorphic in its ability to cause reproductive barrier expression in the cross with T. turgidum and unreduced gamete production in F 1 hybrids . This suggests that some genotypes of Ae. tauschii may have had more opportunities to be involved in the origin of common wheat than others because they frequently hybridized with T. turgidum and because the F 1 hybrids were likely to produce unreduced gametes. In fact, artificial cross-studies have shown that southern Caspian accessions of the TauL2 lineage have a high potential for natural hybridization with T. turgidum (Matsuoka & Takumi, 2017). Thus, the genealogically and geographically structured polymorphism in the crossability with T. turgidum may have had a profound impact on the spatial patterns of common wheat speciation. In addition, Ae. tauschii has six quantitative trait loci (QTLs) for genome doubling in the F 1 hybrids it produces with T. turgidum (Matsuoka et al., 2013). However, very little is known about the natural variation in T. turgidum's ability to cause reproductive barrier expression in crosses with Ae. tauschii. In addition, how variation in the ability to cause genome doubling in F 1 hybrids with Ae. tauschii is structured in this species remains to be addressed. Previous artificial cross-studies have shown that the T. turgidum genotype influences the degrees of viability and fertility in F 1 hybrids with Ae. tauschii (Kihara et al., 1965;Nishikawa, 1953). T. turgidum has a QTL that is involved in unreduced gamete production in the hybrids (Hao et al., 2014). Nevertheless, further studies are required to clarify the role of T. turgidum in the evolution of common wheat and to answer other questions such as: Do cultivated and wild forms differ in reproductive isolation patterns in crosses with Ae. tauschii? Do they differ in the ability to cause hybrid inviability and genome doubling? What is the genetic basis of the ability to cause hybrid genome doubling in this species?
Answers to these questions are essential to understand the genetic underpinning of common wheat evolution.
In the present study, a series of artificial cross-experiments, QTL analysis, and cytological observations were conducted to address the above questions. In the artificial cross-experiment, a diverse array of T. turgidum accessions were crossed with a single Ae. tauschii tester accession with a very short genetic distance to the D genome of common wheat, with the aim to evaluate the natural variation patterns of crossability with Ae. tauschii and the abilities to cause inviability and genome doubling in F 1 hybrids. The QTL analysis and cytological observations were conducted to clarify the genetic basis of the ability of T. turgidum to cause genome doubling in F 1 hybrids with Ae. tauschii.
Major findings were as follows: (a) the cultivars were less crossable with the tester than with the wild accessions were, but the crossability value ranges widely overlapped between the two groups, (b) the proportion of the accessions that produced viable, fertile F 1 plants was similar in cultivars and wild accessions, (c) hybrid genome doubling frequencies were generally increased in cultivar-derived F 1 versus wild-accession-derived F 1 genotypes, (d) some cultivar-derived F 1 genotypes had very high hybrid genome doubling frequencies (>50%), and (e) in the cultivars, a relatively small number of loci that influence unreduced gamete production in F 1 hybrids with Ae. tauschii might be involved in the enhanced ability to cause hybrid genome doubling. On the basis of these findings, we discuss insights into the role of T. turgidum in the evolution of common wheat.

| T. turgidum accessions
This study used 46 cultivated and 31 wild T. turgidum accessions (Table S1) (two accessions) were used. All the cultivated accessions were landraces, except for the modern cultivar, T. turgidum subsp. durum cv. "Langdon" (called LDN here after). Wild accessions (T. turgidum subsp. dicoccoides) were sampled to represent the subspecies distribution range and chloroplast DNA haplotype diversity. Thirty of the 32 chloroplast DNA haplotypes found in a source collection were included in the present study (Mori et al., 2003).

| Ae. tauschii tester accession
The use of an appropriate Ae. tauschii tester is essential for the artificial cross-experiment to be informative with regard to the role of T. turgidum in the evolution of common wheat. In the present study, we used a single Ae. tauschii accession, KU-2103, as the tester. This accession belongs to the TauL2 lineage (Matsuoka et al., 2013), and it was provided by the Plant Germ-plasm Institute of Kyoto University and the National BioResource Project (NBRP)-KOMUGI. We used KU-2103 as the tester because (a) it is one of the accessions that had the shortest genetic distance to the D genome of common wheat (Wang et al., 2013), (b) it has positive alleles at the major QTLs that are involved in the expression of genome doubling in F 1 hybrids with LDN (Matsuoka et al., 2013), and (c) it originated in the southern coastal Caspian region of Iran, which is a candidate region for the cradle of common wheat (Kihara et al., 1965, reported as "stock number 2148" therein). KU-2013 was also used to produce a segregant population for the QTL analysis. The F 1 hybrid seeds were germinated in Petri dishes at 20°C, transplanted into individual pots in early winter, and grown in an unheated greenhouse. Up to six seeds were used per F 1 genotype. Colchicine was not applied to the F 1 plants. Hybridity was confirmed morphologically by comparing coleoptile color, waxiness, and spike shape between the F 1 and their maternal parent plants. For selfing, spikes on well-developed tillers were individually bagged before anthesis. After harvest, well-developed seeds set in the first and second florets of each spike were counted. For each F 1 genotype, the hybrid genome doubling frequency was determined as the selfed seedset rate, as seedset by selfing is a good indicator of the occurrence of genome doubling via unreduced gamete fusion (Kihara, 1946). The hybrid genome doubling frequency was calculated for each F 1 genotype as the number of seeds set divided by the number of florets examined for seedset.

| Statistical analysis
Differences between the cultivated and the wild T. turgidum accessions in crossability with the KU-2103 tester and in their ability to cause genome doubling in the F 1 hybrids were evaluated using a model selection approach based on generalized linear mixed model (GLMM) analysis. In this analysis, an alternative model that included a fixed effect was compared with a null model without a specified fixed effect, using the lme4 package (Bates et al., 2015) for R ver. 3.6 (R Core Team, 2019). Both models were generated using the glmer function of the package, which fits a fixed effect and random effects with a binomial error distribution on a logit link function based on a maximum likelihood estimation method. Model selection was performed based on the minimum Akaike Information Criterion (AIC) value, which is an estimate of the amount of information lost when a given model is used to approximate the process that generated the observed data. Overdispersion was estimated using the overdisp_fun function (Bolker, 2019). Departure of the alternative model from the null model was examined by a likelihood ratio test using the anova function. The theoretical marginal R 2 (the proportion of variance explained by the fixed effect) and conditional R 2 (the proportion of variance explained by the fixed effect and random effects) values of the alternative model were calculated using the r.squarredG-LMM function of the MuMIn package (Bartoń, 2019;Nakagawa & Schielzeth, 2013). The R commander package (Fox, 2005) was used to conduct a chi-squared test for the difference between cultivated and wild T. turgidum in terms of the proportions of the accessions that produced inviable F 1 plants.

| QTL analysis
QTL analysis was used to identify loci in the T. turgidum genome that influence the occurrence of genome doubling in F 1 hybrids with Ae. tauschii. A T. turgidum accession that generated an F 1 genotype with a high genome doubling frequency when crossed with the KU-2103 tester (termed "high accession") and an accession that generated an F 1 genotype with a low genome doubling frequency when crossed with the KU-2103 tester (termed "low accession") were selected from the collection. To produce the segregant population, we first crossed the high accession (female) with the low accession (male). F 1 plants of the cross between the high and low accessions were crossed with the KU-2103 tester (male) to obtain seeds of trihaploid segregants (ABD genome) in which only the loci on the A and B genomes were segregated. and rate calculation were conducted as described above for measuring the hybrid genome doubling frequency.
DArTseq examines numerous genome fragments, aligns these fragments to a reference genome sequence, and produces a biallelic single nucleotide polymorphism (SNP) dataset. The SNP alleles were coded as "1" (present) or "0" (absent) and further screened for use based on the following criteria: (a) the SNPs must be physically mapped to the A or B genome of the T. aestivum reference genome sequence provided by Diversity Arrays Technology (wheat_ ChineseSpring04); (b) the SNPs must be homozygotic in all individuals given their haploid nature (i.e., heterozygotic SNPs are considered erroneous); (c) the SNPs must have no missing values in the high-accession-KU-2103 and low-accession-KU-2103 F 1 hybrids, because this information is required for determining the source T. turgidum accessions of the SNP alleles; and (d) the missing data percentage for each SNP must be less than 5%.
A linkage map was then constructed on the basis of the qualified SNPs and the MSTmap algorithm implemented in the ASMap package, using the mstmap function with the "bychr" and "anchor" options (Taylor & Butler, 2017;Wu et al., 2008). Genetic map distances were calculated based on the Kosambi function (Kosambi, 1944). The QTL analysis was carried out using the R/qtl package based on the model for recombinant inbred lines produced by selfing, because the haplotype data for the trihaploid segregants had two categories, with no possibility of heterozygosity . Conditional genotype probabilities were calculated using the calc.genoprob function (density, 1 cM), assuming a genotyping error rate of 0.001. The sim.geno function was used to impute missing genotypes, with 256 draws. Single-QTL analyses were performed using the scanone function based on the EM algorithm (for the binary trait) and multiple imputation (for the quantitative trait). The statistical significance of putative QTLs was examined by 100,000-fold permutation tests.
Approximate 95% Bayesian credible intervals for the QTL locations were calculated using the bayesint function with the "expandtomarkers" option. QTL model fitting was done using the makeqtl and fitqtl functions. In multiple-QTL modeling, the addint function was used to test pairwise interactions among QTLs.

| Cytological observations
Meiotic cell divisions in the pollen mother cells (PMCs) were analyzed by the conventional aceto-carmine squash method (Matsuoka & Nasuda, 2004) with some modifications. Immature anthers were collected from young spikes, fixed in a mixture of anhydrous ethanol and acetic acid (3:1), and stored at 4°C until use. For the tracking of cell division progression, anthers were sampled from consecutive primary or secondary florets arrayed along the rachis of a spike.
For microscopy, fixed anthers were stained with aceto-carmine at 20˚C for several hours and then squashed in a drop of 45% acetic acid. Cells were observed and imaged under a BX-51 microscope equipped with a DP21 digital camera (Olympus, Tokyo, Japan).

| Crossability
The artificial cross-experiment provided the crossability values for the 77 T. turgidum accessions (Figure 1; Table 1; Table S1). The overall median, mean, and standard deviation were 0.05, 0.11, and 0.14, respectively. In the cultivars (46 accessions), the crossability values varied from 0.00 to 0.40, and the median, mean, and standard deviation were 0.04, 0.07, and 0.09, respectively. In the hulled cultivars (T. turgidum subsp. dicoccum), the crossability values ranged from 0.00 to 0.28. Sixteen of the 46 cultivars (37.0%) had a crossability value of 0.00. In the 31 wild accessions, the crossability values varied from 0.00 to 0.56, and the median, mean, and standard deviation F I G U R E 1 Box and dot plots of the crossability values of cultivated and wild forms ofTriticum turgidum. In the cultivated form plot, triangles and circles denote free-threshing and non-freethreshing accessions, respectively were 0.11, 0.17, and 0.16, respectively. Four of the 31 wild accessions (12.9%) had a crossability value of 0.00. In general, crossability with the Ae. tauschii tester tended to be higher in the wild accessions than in the cultivars, but the value ranges overlapped widely between the two groups.
To further examine the difference in crossability between the cultivars and wild accessions, we analyzed the seedset data by GLMM analysis (Table S2). The GLMM used "form" (i.e., cultivar vs. wild) as the fixed effect and "accession" as the random effect.

| Hybrid viability
Thirty-three T. turgidum accessions produced at least one viable F 1 plant in the artificial crosses (Table S1), while 44 accessions produced no viable F 1 plants. Of these, 20 accessions did not set seeds, and 24 accessions set seeds but produced F 1 plants that were not viable due to hybrid abnormalities, such as germination failure or severe dwarfness (Table S1). The proportion of the accessions that set seeds in the artificial crosses but produced abnormal F 1 plants was higher in the wild accessions (15 out of 27 accessions, 55.6%) than in cultivars (nine out of 30 accessions, 30.0%); however, this difference was not statistically significant (chi-squared test, p = 0.05).
The difference in the hybrid genome doubling frequency between the cultivar-and the wild-accession-derived F 1 genotypes was further examined by GLMM analysis of selfed seedset data (Table S3). In the GLMM, "form" (i.e., cultivar-vs. wild-accession-derived) was the fixed effect, whereas "parental T. turgidum accession" and "individual F 1 plant ID" were the random effects. Goodness of prediction was improved in the alternative model (AIC = 1,367.0) relative to that of the null model (AIC = 1,373.2) (

| QTLs for hybrid genome doubling
The hybrid genome doubling frequency spectrum suggested that   and 23.5 cM, respectively ( Figure S1).
The hybrid genome doubling frequency distribution had a spike at 0.00 (Figure 4). Therefore, we first transformed the original trait into a binary trait (defined as "1" when the original phenotype was >0.00 and "0" when the original phenotype was 0.00) and a quantitative trait (the same as the original phenotypes, except that 0.00 was treated as missing), and then separately performed QTL analyses for each binary and quantitative trait (Broman, 2003) (Table S4).

| Male sporogenesis in the KU-9882-KU-2103 F 1 hybrids
Triploid F 1 hybrids between LDN and KU-2103 set seeds because they produced unreduced gametes at a high rate through nonreductional meiosis. During male gamete formation, PMCs undergo a single cell division and produce pollen dyads instead of tetrads (Matsuoka et al., 2013). However, why triploid F 1 hybrids between the low- At prophase, 21 univalent chromosomes were visible (Figure 6a,b). In several cells, the univalents were observed in a few clusters (Figure 6c).
No PMCs displayed univalents that were aligned to the spindle equator. Subsequently, the univalents split into sister chromatids, moved to the multiple spindle poles, and formed restitution nuclei (Figure 6d,e).

| Possible key role of hybrid genome doubling in the evolution of common wheat
In the widely accepted scenario for the origin of common wheat, the cultivated form, not the wild form of T. turgidum, is the female progenitor (Kihara, 1966;Kihara, 1975). This hypothesis is based mainly on the fact that wild T. aestivum has never been found, indicating that, under natural conditions, successful hybridization between wild T. turgidum and Ae. tauschii occurs rarely. Cultivated T. turgidum and Ae. tauschii are known to co-exist in the wheat fields of northern Iran but natural hybridization between these species has never been observed . In a previous series of artificial cross-experiments, wild T. turgidum was often involved in the expression of reproductive barrier phenotypes (such as chlorosis and necrosis) in F 1 hybrids with Ae. tauschii. This suggested that strong reproductive isolation through postzygotic mechanisms might explain why wild T. aestivum does not exist (Kihara, 1966;Kihara, 1975). However, the genetic relationships between the Ae. tauschii testers and the D genome of common wheat were not addressed in those experiments. Thus, the interpretation of the results in terms of understanding the origin of common wheat is not straightforward.
In the present study, we used the KU-2103 tester in the artificial cross-experiment and found that (  %var denotes the estimated proportion of the phenotype variance explained.
na denotes not applicable for this category.
In the binary trait model, LOD score (relative to the no QTL model), %var by all terms in the model, and p-value based on the LOD score were 13.1, 14.9, and 5.1e−13, respectively.
In the quantitative trait model, LOD score (relative to the no QTL model), %var by all terms in the model, and p-value based on the LOD score were 12.6, 16.6, and 1.4e−12, respectively.
influenced the likelihood of successful formation of allohexaploid offspring.
An obvious caveat here is that our artificial cross-experiment relied on a single Ae. tauschii tester. In general, the genotypes of the Ae. tauschii accessions influence the crossability with T. turgidum and the viability and genome doubling frequency of the F 1 hybrids (Matsuoka & Takumi, 2017;. Thus, the phenotypes of these traits may vary in the F 1 hybrids, even among the Ae. tauschii testers that are genetically close to the common wheat D genome. For this reason, further studies that use testers other than KU-2103 are required to evaluate the validity of our inference regarding the importance of hybrid genome doubling in the evolution of common wheat. Additionally, in this regard, we note that results similar to our findings were obtained in the previous studies. For example, Nishikawa (1964) performed an artificial cross-experiment using 11 T. turgidum accessions (one wild and ten cultivated accessions) and five Ae. tauschii accessions (including one artificial autotetraploid accession). Consistent with our findings, the reported mean crossability value (over the five Ae. tauschii testers) was larger for the wild T. turgidum accession (0.334) than for the cultivated accessions (0.073-0.165). In addition, F 1 genotypes derived from three out of the five wild forms grew normally, but did not set F 2 seeds.
Thus, a relatively high crossability with Ae. tauschii and a reduced ability to cause genome doubling in F 1 offspring might be common phenotypes in wild T. turgidum.

| Genetic basis of hybrid genome doubling
Hybrid genome doubling through unreduced gamete fusion is an important mechanism in plant allopolyploid evolution (Ramsey & Schemske, 1998). Therefore, clarifying the mechanical underpinning of genome doubling in the T. turgidum-Ae. tauschii F 1 hybrids is essential to understand the evolution of common wheat. We addressed this issue by conducting QTL analysis. The three QTLs found for each binary and quantitative trait were mapped to similar locations   Matsuoka et al., 2013). Interestingly, the nonreductional meiosis in KU-9882-KU-2103 F 1 clearly differed from that in LDN-KU-2103 F 1 in that the 21 univalents did not align at the spindle equator prior to the restitution phase ( Figure 6c). This suggests that the bipolar spindle structure might not be formed on the univalents of KU-9882-KU-2103 F 1 plants. If this were the case, the QTLs found in the present study might be involved in the formation of the bipolar spindle on the univalents. To date, the meiotic genes that have been functionally characterized in wheat include RecQ-7, which is located on chromosome 2A (Gardiner et al., 2019).
However, this helicase gene is involved in gene conversion in normal meiosis and thus may not be eligible as a candidate gene for the chromosome 2A QTL.
In conclusion, this study suggests that the enhanced ability of cultivated T. turgidum to cause genome doubling in F 1 hybrids with Ae. tauschii might have increased the probability of the emergence of common wheat. It also suggests that this enhanced ability might be the result of alterations in a relatively small number of loci. The nature of such alterations remains obscure, but they may have been mutations at the QTLs after the domestication of T. turgidum.
Alternatively, the alterations might have occurred through the introgression of alleles that already existed in some unfound wild T. turgidum populations. Further studies on those QTLs may increase our understanding of the evolution of common wheat.

ACK N OWLED G M ENTS
We thank Taihachi

CO N FLI C T O F I NTE R E S T
No conflict of interest has been declared by the authors.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that supports the findings of this study are available in the supporting information of this article.