This study primarily included 1,041 accessions of the Aegilops species preserved and maintained in the WGRC gene bank ( Supplementary Material Table S1 ; Figure 1 ). The accessions were originally collected from various sources and sites including the Middle East, Anatolia, East Asia, and northern Africa ( Figure 1 ; Supplementary Material Table S1 ). Accessions comprise 22 different Aegilops species under five sections (Aegilops, Comopyrum, Cylindricum, Sitopsis, and Vertebrata) (Van Slageren, 1994) and Ae. mutica, which is synonymously known as Amblopyrum muticum. For gene bank curation and most part of the population analysis, only those Ae. tauschii accessions that were not in the previous gene bank curation experiment (Singh et al., 2019a) were used. We also used CIMMYT wheat lines and already curated Ae. tauschii lines ( Supplementary Material Table S1 ) for genotyping together with the diploid Aegilops to dissect the genetic relationships among wheat and Aegilops genomes.

Most of these species are self-pollinated and were primarily maintained by single seed descent, with exceptions described below. Ae. speltoides and Ae. mutica are partially out-crossing and were maintained through sib-mating multiple plants. Specifically, Ae. mutica accessions consisted of 54 samples from five out-crossing plants bulked together.

The DNA extraction, GBS library preparation, and sequencing were performed as we described in our earlier studies (Adhikari et al., 2022a) using two enzyme-based GBS (Poland et al., 2012). Only a single plant per accession was sequenced for all species except Ae. mutica, where we sequenced 54 individuals obtained from randomly crossing five plants, because the species is cross-pollinating and it has a low germination rate.

For the de-novo single nucleotide polymorphism (SNP) calling, reads were demultiplexed using sabre (https://github.com/najoshi/sabre) and adapters were trimmed using fastp (Chen et al., 2018). The variants were called using the available reference assemblies of diploid Aegilops and wheat and using mock references generated as described (Melo et al., 2016; Adhikari et al., 2018). For mock references, the raw GBS reads of selected accessions with higher sequence data were used as the reference source. We also ensured that the mock reference represents the sequences of relevant Aegilops species or the genomes [C, D, M, N, S, U, T] for the population to be genotyped. The de-novo variants were called using BCFtools (Li, 2011) and used for initial gene bank curation and population clustering of the whole collection. Then the de-novo variants were also called for some species independently depending on the objectives of the specific analysis ( Supplementary Material Table S2 ). For some species in polyploid lineages, we called variants on a diploid ancestor and, later, the same variants were called in the polyploids using BCFtools (Li, 2011). After calling variants, unless otherwise stated, we filtered loci to keep any variants passing these conditions: minor allele frequency (MAF) >0.01, missing <30%, and heterozygous <10%.

The TASSEL5 GBSv2 pipeline was used for reference-based SNP calling (Glaubitz et al., 2014). For this method, Ae. tauschii reference genome Aet v5.0 (Wang et al., 2021) or Ae. sharonensis (Yu et al., 2022), Ae. speltoides (Avni et al., 2022), Ae. searsii, and Ae. bicornis (Li et al., 2022) genomes were used. We also called variants in all these diploids species to the wheat reference using the “Chinese Spring” wheat reference (IWGSC CS RefSeq v2.1) (Zhu et al., 2021) to observe the relationship between Aegilops and wheat.

In the first step, the germplasm curation identified misclassified accessions and corrected the taxonomy of these accessions in the database (Singh et al., 2019a). We identified misclassified accessions by constructing a phylogenetic cluster colored with the recorded species. These were further verified using PCA clustering followed by a visual assessment of seeds and spikes. The misclassified accessions were identified and confirmed with multiple genotyping sets viz. entire collection, species alone, and same genome accessions together.

In the second step, the genetically identical accessions were determined using allele matching (Singh et al., 2019a; Adhikari et al., 2022a). However, this assessment was done only for the accessions of the species whose reference genome is available, for example, Ae. tauschii and the Sitopsis section Aegilops. The allele matching (>99% identity by state) was used as a threshold to confirm genetically identical accessions. Allele matching used homozygous and non-missing sites between two given accessions, and the raw markers were filtered using MAF >0.01, missing <50%, and heterozygous <20% parameters before allele matching. We conducted further examinations of the sets of genetic duplicates to assess their phenotypic similarities, collection sites, and sources of collection.

The genotyping matrices were analyzed for the genetic distances among the Aegilops populations, which were then used for exploring the population structure and ancestry. For phylogenetic clustering, the genetic distance was computed using the “dist” function in R (R Core Team, 2020), and the R packages ape (Paradis and Schliep, 2019) and phyclust (Chen, 2011) were then used to generate unrooted neighbor-joining (NJ) tree with the default parameters (Singh et al., 2019b; Adhikari et al., 2022a).

The genetic relationships among the Aegilops accessions were further examined via PCA, which was performed in two steps. The A matrix was derived from A.mat() function within the R package rrBLUP (Endelman, 2011), and the eigenvalues and eigenvectors were derived using the “e” function (Adhikari et al., 2022a). Furthermore, the population structure of the Sitopsis group of Aegilops was also performed with the reference-based genotyping profile using fastStructure software (Raj et al., 2014) as explained (Adhikari et al., 2022a). We computed Nei’s diversity index (Nei, 1987) and total segregating loci for each of the Aegilops species to assess the relative diversity of the species.

We investigated the traditional genome symbols of Ae. columnaris (UM) and Ae. neglecta (UM, UMN) for the presence/absence of the M genome. There are recent cytology-based findings that have questioned the traditional genome symbols of these species (Badaeva et al., 2018). To test this, we computed the sequence read mapping and segregating loci on the M and U mock reference genomes for the Ae. columnaris and Ae. neglecta accessions as well as two other tetraploids (Ae. nelglecta and biuncialis) whose genomic compositions are unequivocally recognized as MU or UM. The de-novo variants were first identified for the diploid M genome (Ae. comosa) and U genome (Ae. umbellulata) populations separately, and then the same variants were called on these four tetraploid species. We also constructed the phylogenetic clustering among Ae. columnaris, Ae. neglecta, Ae. geniculata, Ae. biuncialis, and a tetraploid that shares only the U genome (Ae. triuncialis) to see their relative positions in the tree.

We mapped diploid Aegilops GBS reads to the wheat genome (CS.Ref.v1) (Appels et al., 2018) and computed sequence read mapping coverage. The reads mapped per Mb wheat subgenome and the total variants mapped for each wheat subgenome (A, B, D) were recorded. We did not further evaluate Ae. tauschii whose close genetic relationship as the wheat D subgenome donor has been clearly established. We also generated an unrooted NJ phylogenetic tree among diploid Aegilops and wheat using the variants called on wheat B and D reference subgenomes independently.

Aegilops species characterized in this study were primarily collected around the Fertile Crescent, Anatolia, central Asia, northern Africa, and southern Europe ( Figure 1 ; Supplementary Material Table S1 ). Of the five sections, the Aegilops section [Ae. umbellulata (U), Ae. kotschyi (US), Ae. peregrina (US), Ae. triuncialis (CU), Ae. columnaris (UM), Ae. biuncialis (UM), Ae. neglecta (UM, UMN), Ae. geniculata (MU)] exhibited a much wider distribution from central Asia to northern Africa ( Figure 1 ). The species of Cylindropyrum [Ae. markgraffii (C), Ae. caudata (C), and Ae. cylindrica (CD)] were primarily collected from Uzbekistan, Tajikistan, Kazakhstan, Azerbaijan, and Turkey. The species of the Comopyrum [Ae. comosa (M), Ae. uniaristata (N)] mainly come from Greece, Turkey, and Russia. The Sitopsis (S genome) species [Ae. bicornis, Ae. searsii, Ae. sharonesis, Ae. longissima, and Ae. speltoides] were predominantly collected in Turkey, Israel, Syria, Iraq, and Jordan. The Vertebrata section species [Ae. tauschii (D), Ae. crassa (DM, DDM), Ae. ventricosa (DN), Ae. juvenalis (DMU), and Ae. vavilovii (DMS)] were obtained from central Asia to southern Europe ( Figure 1 ; Supplementary Material Table S1 ). The Ae. mutica tested here originated from Turkey and Armenia ( Supplementary Material Table S1 ).

We identified 54,667 de novo called SNPs for the entire Aegilops collections genotyped together. After filtering (MAF >0.01, missing <30%, and heterozygosity <10%), we retained 46,879 SNPs ( Table 1 ). We removed 10 accessions (TA2674, TA2633, TA1733, TA11097, TA1740, TA2178, TA2042, TA1739, TA2316, and TA2296) with high rate of missing call (>80%). When we separated the genotyping information per species, we identified filtered segregating SNPs in the range of 1,483 for Ae. searsii to 14,322 for Ae. speltoides ( Table 1 ). We also generated other SNP-genotyping matrices for analysis-specific purposes, such as for particular species’ genetic relations and for genetically identical accession determination ( Supplementary Material Table S2 ).

The unrooted NJ phylogenetic tree with all tested Aegilops accessions gave clear separation of species as the branches of clades and sub-clades differentiated all 23 species and the relevant groups ( Figure 3 ). We observed the species sharing genomes as closely related clades, such as Ae. kotschyi and Ae. peregrina (SU) and Ae. geniculata and Ae. biuncialis (UM), clustered into respective primary clades. Overall, there were three primary clades: (i) the first clade consisted of Ae. speltoides and Ae. mutica; (ii) the second clade has four diploids of Sitopsis (except Ae. speltoides), Ae. tauschii, and D genome polyploids (except Ae. cylindrica); (iii) the third primary clade has all other species, including M, N, C, and U genome diploids and polyploids.

The hexaploid (6X) and tetraploid (4X) species within a clade, such as Ae. neglecta and Ae. crassa, were grouped separately by ploidy. The ploidy levels of these genetically clustered sub-groups (6X and 4X) were also verified using chromosome counting ( Supplementary Material Figure S1 ) following Koo et al. (2017). The chromosome numbers of some accessions of Ae. crassa ( Supplementary Material Figure S2 ) were also confirmed with the published data (Badaeva et al., 1998).

PCA also grouped the Aegilops species commensurate with the phylogenetic analysis. The first and second principal components (PC1 and PC2) explained about 17% and 14% of the variations among the Aegilops, respectively. PC1 separated Ae. speltoides from other polyploids and diploids ( Figure 4 ), while the PC2 primarily differentiated Ae. tauschii and Ae. speltoides, the D genome donor to wheat and the potential sister group of the wheat B genome donor, respectively. As in phylogenetic analysis, PCA grouping also divided the 4X and 6X accessions of the Ae. neglecta and Ae. crassa ( Figure 4 ).

As we observed the separation of four Sitopsis members with Ae. speltoides and Ae. mutica, we separately examined the population of these species using reference-based variants from the Ae. speltoides genome assembly. The constructed phylogenetic tree distinctly divided the S-genome diploids into two large clades, one representing Ae. speltoides and the other with the remaining four Sitopsis ( Figure 5 ). The genetic clustering corresponded to the historical sub-section division of the section is Truncata (Ae. speltoides) and the Emarginata. We also observed that the Ae. mutica (T genome) clustered closer to Ae. speltoides both in PCA and phylogenetic analysis ( Figure 5 ). The relationships among Sitopsis group and Ae. mutica were further verified by computing pairwise Nei’s F_ST (Nei, 1987), where we observed Ae. mutica has the closest genetic relationship [lowest F_ST (0.65)] with Ae. speltoides, closer than any other members of the Sitopsis ( Supplementary Material Table S4 ). Hence, all these analyses support that Ae. mutica as the sister taxon to Ae. speltoides and it is an Aegilops species.

Furthermore, within the S-genome diploids, the Ae. speltoides and Ae. searsii had the most genetic differentiation with the highest F_ST value 0.88 ( Supplementary Material Table S4 ). However, the pairwise F_ST indicated that speltoides is genetically almost equally and highly differentiated from all other S-genome diploids (Emarginata) ( Supplementary Material Table S4 ).

Population structure analysis of S-genome diploids matched with the phylogenetic tree and pairwise F_ST analysis. At K = 2, there was a differentiation between Ae. speltoides and the rest of the Sitopsis, while at K = 3, Ae. searsii also differentiated from the rest of the Sitopsis ( Figure 6 ). At K = 7, Ae. bicornis accessions separated from others and then no new differentiation was observed until K = 12. Both in the phylogenetic tree and in population structure analysis, the Ae. longissima and Ae. sharonensis appeared as highly genetically similar groups ( Figures 5 , 6 ). In fact, there was no population differentiation between these two species at any level of K. The pairwise F_ST values also confirmed that these two species have the lowest pairwise F_ST = 0.006 ( Supplementary Material Table S4 ), and the population differentiation is very low. Furthermore, two sub-groups within Ae. speltoides, var. speltoides, and var. ligustica also did not differentiate at any levels of K in the population structure analysis ( Figure 6 ) and the PCA ( Supplementary Material Figure S3 ). However, within Ae. speltoides, a few admixtures were observed and were differentiated for their geographical origins ( Figure 6 ).

Most of the tetraploid Aegilops have the U genome; therefore, understanding the genetic relationship among members of the U-genome clade gives insight into a large set of taxa in the genus. Phylogenetic clustering of these species only showed two larger clades, where one was represented by Ae. triuncialis (UC) and the other had all remaining tetraploids ( Figure 7 ). The diploid Ae. umbellulata sits on the intermediate position between the larger clades. Although the variants were only called on U-genome (Ae. umbellulata) de-novo reference, the tetraploids distinctly grouped for their genomic compositions. The tetraploid species Ae. pregerina and Ae. kotschyi (US genome), Ae. neglecta and Ae. columnaris (traditionally assigned as UM), and the UM genome tetraploids Ae. biuncialis and Ae. geniculata formed a separate clade and sub-clades ( Figure 7 ). Also, we observed the splitting of Ae. umbellulata accessions into smaller clades. With a few exceptions as noted below, these phylogenies largely agree with previous genome designations.

Ae. columnaris and Ae. neglecta formed a different clade than the other tetraploids with U and M genomes such as Ae. geniculata (UM) and Ae. biuncialis (MU) in both phylogenetic clustering and PCA ( Figures 3 , 4 , 7 ; Supplementary Material Figure S4 ). The comparative positions of these tetraploids with other tetraploids in the genetic cluster indicated that these two tetraploids must be given unique genome symbols than the Ae. geniculata and Ae. biuncialis ( Supplementary Material Figure S4 ). Thus, we hypothesized that Ae. columnaris and Ae. neglecta do not carry the M genome. The absence of M genome in Ae. columnaris and Ae. neglecta accessions was further confirmed by computing total reads mapped and total variants called on M-genome (Ae. comosa mock reference) and U genome (Ae. umbellulata mock reference) ( Supplementary Material Figure S5 , Supplementary Material Table S5 ). All four tetraploid species, namely, Ae. columnaris and Ae. neglecta along with Ae. geniculata and Ae. biuncialis exhibited an equal percentage of overall reads alignment (~38%) on the U genome, whereas the percentage read alignment of Ae. columnaris and Ae. neglecta on M genome was low (~21%) as compared to the alignment of Ae. geniculata and Ae. biuncialis reads (~38%). We also noticed that a few Ae. comosa segregating loci were mapped for Ae. columnaris (10%) and Ae. neglecta (24%) on the M genome. In contrast, Ae. biuncialis had 50% and Ae. geniculata had 46% M-genome loci. Hence, the proportion of mapped reads and loci also suggested that the Ae. neglecta and Ae. columnaris must have the U genome, but a different second sub-genome than M. Thus, we proposed that Ae. columnaris and Ae. neglecta genome formulas are most likely UX (X, the unknown genome) or UXN in hexaploid form as proposed based on the cytology (Dvorak, 1998; Badaeva et al., 2018).

For the entire collection, we obtained 54,667 SNPs, which were skewed to low MAF as expected for a diverse population like this ( Supplementary Material Figure S6 ). Despite the differences in population size, the total segregating loci for the species or groups were mostly dependent on the ploidy levels and the reproductive biology (inbreed vs. outcrossing) ( Table 1 ). The polyploids and out-crossing species had a higher number of segregating loci compared to other diploids ( Table 1 ). Notably, the MAF of the loci in partially cross-pollinated species, such as Ae. speltoides, had a higher frequency ( Supplementary Material Figure S7 ) than that of the MAF of the loci for the entire Aegilops collection ( Supplementary Material Figure S6 ).

The Nei’s diversity indices also followed the pattern of segregating loci which were greater in polyploid and cross-pollinated species. We computed Nei’s diversity index for the entire collection as 0.10 ( Table 1 ). Of all 23 species, Ae. bicornis had the lowest Nei’s diversity index (0.012) followed by Ae. searsii (0.013) and Ae. umbellulata (0.015). Among the diploids, the Ae. speltoides had the highest Nei’s diversity (0.072), which was followed by Ae. mutica (0.053). Among the tetraploids, the Ae. triuncialis had the lowest diversity index (0.032) while the Ae. neglecta had the highest diversity index (0.062). The hexaploid species Ae. vavilovii has the highest Nei’s diversity index value among all 23 species analyzed in the experiment ( Table 1 ). This increased diversity can be attributed to various factors such as multiple gene copies, hybridization during speciation, increased mutation rates, and more opportunities for recombination due to the presence of multiple genomes.

The genetic clustering between wheat and all diploid Aegilops showed that Ae. tauschii is the nearest extant Aegilops to the bread wheat ( Supplementary Material Figure S8 ). The genetic cluster clearly showed that Ae. speltoides is not closer to wheat as Ae. tauschii and even other diploids, and supporting that, Ae. speltoides is likely not the direct progenitor of the wheat subgenome B ( Supplementary Figure S8 ). However, the Ae. speltoides read depth mapping and SNP detection occurred at its maximal on the wheat subgenome B ( Figure 8 ), indicating the species as the sister group of wheat B genome progenitor. Furthermore, the other members of the Sitopsis group clustered between Ae. speltoides clade and the clade with Ae. tauschii and the wheat subclades in the phylogenetic tree ( Supplementary Material Figure S8 ). Consistent with the genetic clustering, their maximum read mapping and SNP detection also occurred at subgenome D and B chromosomes ( Supplementary Material Figures S8 – S10 ), suggesting that the four members of Sitopsis, except Ae. speltoides, have very strong genomic relationships with both D and B subgenomes.

Similarly, in the U genome diploid (Ae. umbellulata), the highest proportion of sequence reads was mapped onto wheat chromosomes of the D subgenome, followed by those of the A and B subgenomes ( Supplementary Material Figure S11 ). Exceptionally, a slightly higher proportion of reads were mapped on 2A than the 2D. The pattern of SNP detection was exactly the same as read mapping, indicating that wheat subgenome D is the closest to the U genome of the Aegilops. However, relations between the wheat A genome and the Aegilops U genome cannot be overlooked, as reasonably higher reads and loci were mapped on the A genome as compared to the wheat B genome ( Supplementary Material Figure S11 ). Likewise, the highest number of reads and SNPs were mapped onto wheat subgenome D for the N genome diploid (Ae. uniaristata) ( Supplementary Material Figure S12 ), for the M genome diploid (Ae. comosa) ( Supplementary Material Figure S13 ), and C genome diploid (Ae. markgraffii) ( Supplementary Material Figure S14 ). These observations suggest that the N, M, and C genomes of Aegilops are also genetically closer to the D subgenome than A and B.

Interestingly, the Ae. mutica accessions when mapped onto the wheat subgenomes showed higher sequence read and loci mapped on the wheat D subgenome ( Supplementary Material Figure S15 ). The read and loci mapping pattern was unchanged even when we replaced wheat D subgenome chromosomes with Ae. tauschii chromosomes. Nevertheless, all types of population grouping within Aegilops ( Figures 3 – 5 ; Supplementary Material Figure S8 ) evidently showed that Ae. mutica is a sister group of Ae. speltoides and still a member of B lineage. Some recent studies based on whole genome sequencing data have also reported a higher sequence read and loci mapping of Ae. mutica on the wheat D subgenome compared to others (Grewal et al., 2022; Li et al., 2022).

In this study, we genotyped over a thousand accessions representing almost all species of the Aegilops genus, covering the full range of their natural distributions under the Van Slageren (1994) nomenclature, with missing only Ae. caudata. We curated the WGRC gene bank Aegilops collection, giving curated germplasm sets that are ready to screen for the novel alleles and utilize in the breeding program. The misclassified accession were confirmed with multiple analyses including phylogenetic clustering of the whole population, species or genome-specific populations and PCA, therefore there is strong support for the genotype-based identification of these misclassified accessions ( Supplementary Material Table S3 ). Since the genotype-based clustering evidently differentiated the hexaploid and tetraploid accessions within the species such as Ae. crassa and Ae. neglecta, we can also provide the ploidy levels information as a means of within-species classification and update the gene bank database.

Here, we identified the redundant accessions in the species with variants called directly on reference genome assemblies. This gives increased power and accuracy in variant calling. Therefore, we suggest the re-assessment of genetically redundant accessions for other Aegilops species in the future when reference assemblies are available. For the polyploid Aegilops, reference variant calling can be done whenever the component species reference genomes are available using a combined reference genome or independent variant calling to each genome. As we examined the origins of these genetically verified and visually confirmed duplicates, we discovered that many of them originated from various research institutes rather than directly from collectors. Therefore, we here recommend the need for curating the global collection of these naturally collected germplasms, as the same genetic materials can be preserved under different plant IDs or accession numbers. In our previous studies, we also observed several duplicates originating from the exact same collection sites (Singh et al., 2019a; Adhikari et al., 2022a). This is because these self-pollinated species have already reached genomic saturation, and the progeny of the same mother parents are genetically identical inbred. Although we do not suggest discarding the duplicated accessions identified here, we strongly suggest for considering these results when utilizing the collection, such as screening the accessions for disease resistance or developing introgression populations. Overall, gene bank curation helps in the management, preservation, and utilization of the germplasms (Singh et al., 2019a; Volk et al., 2021).

This is the most comprehensive Aegilops population genetic study reported so far with over 45 thousand de-novo filtered SNPs and reference-based variants. In the study, we took advantage of recently completed chromosome-scale genome assemblies of diploid Aegilops (Wang et al., 2021; Avni et al., 2022; Li et al., 2022; Yu et al., 2022). Until now, the lack of genomic resources including reference assemblies has been a major issue hindering the species population genomic analysis. Therefore, future genomic studies on Aegilops must focus on generating more genomic resources for other diploids and polyploids. With a larger population and thousands of genomic variants, the population grouping that we observed here was at the finest level, enabling us to differentiate the 4X and 6X accessions within a species ( Supplementary Material Figure S1 ).

Our genetic analysis supports that the Ae. mutica requires no genus-level separation from other Aegilops as Van Slageren (1994) suggested. It is genetically an Aegilops taxon closer to Ae. speltoides ( Figures 4 , 5 ). This is in agreement with recent reports (Bernhardt et al., 2020; Li et al., 2022). Further genomic analysis may require high coverage genomic data and a greater number of samples to better understand the relationship among Ae. mutica and other diploid Aegilops. Additionally, the genetic differences that we observed here between the Truncata (Ae. speltoides) and Emarginata (four other) Sitopsis were greater; therefore, the redefinition of the section Sitopsis could be desirable. One of the ideas could be the separation of Ae. speltoides from the rest of the four Sitopsis members and regrouping the Ae. speltoides with Ae. mutica ( Figures 3 – 5 ; Supplementary Material Figure S8 ).

We also showed that the Ae. sharonensis and Ae. longissima have very high genetic similarities or a low genetic differentiation (F_ST = 0.006) and are most likely the sub-species of the same species. Also, both of these species are equally distant from Ae. speltoides. The finding is also supported by the latest study, where Avni et al. (2022) reported that the genomes of these two species are highly similar with identical genome sizes and also share 292 orthogroups.

In this study, we observed a little genetic difference between the two sub-taxa of Ae. speltodies; var. speltoides and ligustica with no population differentiation ( Figure 6 ; Supplementary Material Figure S3 ), in accordance with several past studies. These two sub-groups of speltoides not only have distinct spike morphology and mode of seed dispersal but also exhibit similar karyotype structure, producing fully fertile hybrid and mixed stands of two types naturally exhibits (Zohary and Imber, 1963). A single locus Lig on chromosome 3S governs the spike morphology of these two sub-groups (Luo et al., 2005); otherwise, they are highly genetically similar.

The U genome tetraploids and its progenitor Ae. umbellulata genetic clustering revealed the unique relationships among the species. We observed the Ae. umbellulata accessions split into sub-groups in such a way that some accessions were clustered closer to Ae. triuncialis clade whereas some other accessions reposed near the other tetraploid clades ( Figure 7 ), suggesting the potential unique Ae. umbellulata ancestries for the two groups.

In this study, we found further evidence that the Ae. columnaris and Ae. neglecta genome symbols should not include the M genome designation ( Supplementary Material Figures S4 , S5 and Supplementary Table S5 ), based on sequence read and loci mapping data, and phylogenetic clustering ( Supplementary Material Figure S4 ). Cytology-based approaches (Resta et al., 1996; Dvorak, 1998; Badaeva et al., 2004; Badaeva et al., 2018) have previously discussed this issue and suggested the symbol “X” (Resta et al., 1996). Several lines of evidence, including low chromosome pairing in hybrids of Ae. columnaris x Ae. comosa (the M genome progenitor), variation in repetitive nucleotide sequences, and differences in the karyotype structure C-banding pattern, have been used to confirm the absence of the M genome in Ae. neglecta and Ae. columnaris (Badaeva et al., 2018). This study has provided further verification with thousands of loci. Therefore, we suggest research communities for the consistent use of genome symbols for Ae. columnaris (UX) and Ae. neglecta (UX or UXN). Furthermore, cytological and genomic evaluation of the X genome is certainly warranted.

Ploidy level and the mode of fertilization appeared as major determinants of Aegilops accessions diversity ( Table 1 ). Interestingly, we did not observe the direct impact of population size on Nei’s diversity index (Nei, 1987) at any ploidy levels ( Table 1 ). For example, the diploid Ae. sharonensis (nine accessions) exhibited a higher diversity index (0.019) compared to Ae. umbellulata (58 accessions), and the tetraploid Ae. ventricosa (17 accessions) had a higher diversity index than another tetraploid, Ae. triuncialis (199 accessions) ( Table 1 ). Additionally, we noted that Ae. speltoides, as the diploid species, displayed the greatest diversity, and relatively higher diversity indices were observed in the S genome polyploids such as Ae. kotschyi, Ae. peregrina, and Ae. vavilovii ( Table 1 ). In summary, most of the Aegilops species exhibited a wider and more variable diversity and had greater potential to be utilized in wheat breeding. Therefore, it is crucial to make serious efforts toward the in-situ conservation of these germplasms and enhance ex-situ Aegilops germplasm collections. Kilian et al. (2011) also emphasized the urgency of protecting these Aegilops germplasms, highlighting the importance of understanding Aegilops genetic diversity, Aegilops-Triticum molecular biological relationships, and identifying and preserving suitable Aegilops alleles for wheat breeding.

This study represents, perhaps, the first comprehensive report on genomic relationships between all Aegilops genomes and wheat sub-genomes, based on high-throughput sequence-based markers and robust phylogeny of these wild wheat species. Consistent with some earlier reports, our findings indicate that most of the Aegilops genomes (U, M, N, C) are genetically closer to the wheat D subgenome ( Supplementary Material Figures S9 - S15 ), with the exception of Ae. speltoides ( Figure 8 ). Several studies have reported that the speciation event of the B genome donor occurred earlier than the speciation of Ae. tauschii (the D-genome lineage), resulting in stronger evolutionary relationships of the U, M, N, and C diploid Aegilops within the D-genome lineage (Glémin et al., 2019; Tanaka et al., 2020; Said et al., 2021).

In our study, we observed unique relationships between certain genomes within the Aegilops-Triticum complex that had not been clearly described in earlier studies. One of the most important observations is that four Sitopsis species exhibit relationships with both the B and D subgenomes of wheat. These relationships were evident in the phylogenetic tree and supported by statistic on sequence read and mapped loci ( Supplementary Material Figures S8 - S10 ). Interestingly, recent reports have also considered these four Sitopsis members as part of the D lineage, and are closer to the wheat D subgenome (Li, 2011; Avni et al., 2022; Li et al., 2022).

In this study, we observed unique genetic characteristics of Ae. mutica as it was phylogenetically closer to the Ae. speltoides ( Figures 3 – 5 and Supplementary Material Figure S8 ); however, it showed genetic similarities with the wheat D subgenome ( Supplementary Material Figure S15 ). Interestingly, similar observations have been reported in recent studies. Li et al. (2022) reported lower genetic similarities between Ae. mutica and wheat B subgenome computed as genetic relatedness. Likewise, Grewal et al. (2022) reported a similar relationship between Ae. mutica and wheat subgenomes, with the highest number of Ae. mutica loci mapped on the D subgenome, rather than the A and B subgenomes ( Supplementary Material Figure S15 ). Therefore, the genetic similarities and phylogenetic relationship between the Ae. mutica and the Aegilops-Triticum complex are exclusive and warrant further investigation in a larger population with high-depth sequencing. Furthermore, these analyses indicate that Ae. mutica genome may have undergone independent evolution or played a role in the evolution of polyploid genomes following its divergence from Ae. speltoides. Some recent studies also argued that Ae. mutica and the D lineage underwent homoploid hybridization followed by introgression (Bernhardt et al., 2020; Li et al., 2022). Bernhardt et al. (2020) reported that most of the members of the Aegilops genus, except Ae. speltoides, likely evolved through ancient primordial hybrid speciation events involving the ancestral Triticum and Ae. mutica. Earlier studies also indicated a higher degree of homology between Ae. mutica and the wheat D subgenome (Jones and Majisu, 1968).

This study establishes a solid foundation for the future utilization of Aegilops germplasm within the WGRC gene bank. The development of introgression populations, combined with new genomic tools, has the potential to accelerate the selection and advancement of novel alleles in wheat breeding. In an ongoing investigation, we have successfully created wheat—Ae. speltoides introgression lines and have achieved the mapping of introgression segments using a skim-sequencing approach (Adhikari et al., 2022b). Likewise, association genomics approaches can be leveraged to identify novel Aegilops alleles directly within the wild germplasm collections (Gaurav et al., 2022). As an example, candidate genes associated with various agronomic traits in another wild wheat relative, einkorn, were identified using the cost-effective skim-sequencing technique (Saripalli et al., 2023). Within this context, the importance of these highly diverse Aegilops accessions is further enhanced. Finding trait-related alleles through genome-wide association studies, generating reference assemblies, and resequencing diverse panels represent some of the future steps in harnessing the potential of these valuable Aegilops genetic resources for enhancing wheat.

In conclusion, this study has unveiled the genomic and genetic relationships among all Aegilops species and demonstrated the efficient use of the GBS approach for curating gene bank accessions and investigating the genetic diversity and population structure of the entire Aegilops collection. Most likely this is the first genomic analysis of a nearly complete set of the genus Aegilops encompassing 23 species. We dissected a larger population (1,041) using over 45K SNPs and constructed a robust phylogenetic tree and the PCA clusters. The population grouping and structuring of this valuable wild wheat species largely align with the traditional nomenclatures at the species level. Moreover, using these high-throughput genome-wide markers, we have confirmed the genome symbols of two tetraploid species that were previously under debate in the literature.

Our findings also reveal that each Aegilops subgenome and wheat subgenomes exhibit unique relationships at the genomic level, warranting further investigation. Notably, Ae. mutica showed unique characteristics, appearing as a sister group of Ae. speltoides, yet displaying a higher number of sequences and variants mapped onto the wheat subgenome D. The genetic and evolutionary relationships among Aegilops and with wheat will become clearer when we have more genomic resources, such as genome assemblies and resequencing data for each Aegilops species. This study offers a comprehensive view of the relative genetic diversities of all 23 species together for the first time. The substantial genetic diversity observed, along with its relative extent in each Aegilops species, presents an opportunity to select species and germplasms as sources of novel alleles for wheat breeding and improvement.