Chromosome and molecular analyses reveal significant intraspecific karyotype diversity and provide new evidence on the origin of tetraploid grass Aegilops columnaris

Aegilops columnaris Zhuk. is tetraploid grass species (2n=4x=28, UcUcXcXc) closely related to Ae. neglecta and growing in Western Asia and a western part of the Fertile Crescent. Genetic diversity of Ae. columnaris was assessed using C-banding, FISH, nuclear and chloroplast (cp)DNA analyses, and gliadin electrophoresis. Cytogenetically Ae. columnaris was subdivided into two groups, C-I and C-II, showing different karyotype structure, C-banding and FISH patterns. Group C-I was more similar to Ae. neglecta. All types of markers revealed significant heterogeneity of the C-II group, although group C-I was also polymorphic. Two chromosomal groups were consistent with plastogroups identified in a current study based on sequencing of three chloroplast intergenic spacer regions. The similarity of group C-I of Ae. columnaris with Ae. neglecta and their distinctness from C-II indicate that divergence of the C-I group was associated with minor genome modifications. Group C-II could emerge from C-I relatively recently, probably due to introgression from another Aegilops species followed by a reorganization of the parental genomes. Most C-II accessions were collected from the very narrow geographic region, and they might originate from a common ancestor. We suggest that the C-II group is at the initial stage of species divergence and undergoing an extensive speciation process.


Introduction
Aegilops columnaris Zhuk. is annual tetraploid (2n=4x=28) grass species naturally growing in Western Asia, mainly in Turkey, Armenia, and in a western part of the Fertile Crescent [1][2][3]. It is also native to Crete, Iraq, Lebanon, Azerbaijan, and Iran but found as adventive species in France, near Marseille [1]. Despite a relatively broad distribution area, Ae. columnaris is uncommon throughout its range. Biodiversity Collecting Mission Database included 816 Ae. columnaris site records (https://www.gbif.org/), and according to Genesys, 763 accessions are currently maintained in gene banks worldwide (https://www.genesys-pgr.org/). This number, however, can be overestimated owing to a large number of potentially duplicated and incorrectly classified accessions. On another side, many new sites were recently discovered during collection missions. However, the novel samples (e.g., reported in [4] or materials analyzed in a current study) were not included in these databases.
The source of the second genome of Ae. columnaris and Ae. neglecta is still unknown. H. Kihara [17] suggested that it could be related to the M-genome of Ae. comosa Sm. in Sibth. & Sm. based on morphological comparisons and analysis of meiotic chromosome pairing in Ae. columnaris x Ae. biuncialis (2n=4x=28, UUMM) hybrids. He designated this genome as "modified M," and this symbol is still used in most taxonomical systems [1,2,10,16,17,19,25]. However, the F1 hybrids of Ae. columnaris x Ae. comosa exhibited low chromosome pairing [15]. Differences in the patterns of variation of the repetitive nucleotide sequences [11,26], RAPD-spectra [9], the results of DArTseqbased analysis [12], comparison of karyotype structures [19], C-banding and FISHpatterns [13,20] contradicted this hypothesis. Taking into consideration the distinctness of Ae. columnaris genomes J. Dvořák [26] suggested to change its genome formula from the UM to UX 1 . More recent data of DArTseq-based analysis revealed higher similarity of the second genome of Ae. columnaris and Ae. neglecta with the genome of Ae. speltoides Tausch or Ae. mutica Boiss. [12], therefore, a new genomic formula, UT s, was proposed for these tetraploid species.
In the previous publication [13], we uncovered the significant karyotype diversity of Ae. columnaris, which was expressed in a variation of the C-banding patterns and, despite a small number of accessions studied, translocation polymorphism. In this paper, however, the translocations were classified tentatively due to the lack of standard genetic nomenclature of Ae. columnaris chromosomes. The problem of chromosome classification was solved later when a set of wheat-Ae. columnaris (K-1193) introgression lines was developed and cytogenetically characterized [20].
These introgression lines also enabled the identification of gliadin components encoded by particular Ae. columnaris chromosomes [27]. Although extensive polymorphism of electrophoretic spectra of gliadins was demonstrated for durum and common wheat [28], these markers are broadly exploited for wheat cultivar identification [29] diversity of gliadin profiles of Aegilops species, including Ae. columnaris, is much lesser studied. Publications were mainly focused on Ae. tauschii, the D-genome donor of common wheat [30,31], and only a few papers described other Aegilops species [32][33][34].
The aim of the present study was the analysis of intraspecific diversity of Ae. columnaris on a broader sample of accessions using cytogenetic (C-banding, FISH with various DNA probes), biochemical (seed storage proteins -gliadins), and molecular (comparative sequence analysis of nuclear and chloroplast DNA fragments) markers.

C-banding analysis of Ae. columnaris
We showed that most of Ae. columnaris accessions were karyotypically uniform, but two, K-4224 and a sample provided by Drs. E.A. Nazarova and A.G. Gukasyan in 1998 consisted each of three distinct biotypes, while PI 554187 -of two biotypes (Table 1). Three accessions (PI 564182 from Turkey and K-4553 and K-4233 from Armenia) maintained in gene banks under the name Ae. columnaris were found to be taxonomically misclassified: and were indeed Ae. neglecta. Accession IG 49067 was the mix of Ae. columnaris and Ae. biuncialis, whereas accessions K-2344 from Armenia and AE 1607 of unknown origin -the mix of Ae. columnaris and Ae. triuncialis. One K-4224 genotype was the F1 hybrid between Ae. columnaris and Ae. triuncialis (Fig. 1, c). Chromosomes are designated according to genetic nomenclature; the U c / U tr chromosomes are shown in dark green, the X c /X tr chromosomes in dark blue, the C t of Ae. triuncialis -in red, and the U t in light-green color). Red arrowheads point to satellite chromosomes. Blue arrows show translocated 5U c :6X c chromosomes (a).
The C-banding method revealed that Ae. columnaris gene pool consists of two distinct karyotypic groups that differed from each other and from Ae. neglecta ( Fig. 1, a, b, d) in karyotype structure, the number of satellite chromosomes, and C-banding patterns. The first, larger group was designated C-I, whereas the second, smaller one -C-II. Group C-I included 62 accessions collected from an entire distribution range, while C-II comprised only seven accessions -six from the southeast coastal part of Turkey and one from Iraq (Fig. 2). Each Ae. columnaris group demonstrated high diversity of the Cbanding patterns and, in the case of C-I, broad translocation polymorphism (Figs. 3-7). Thus, karyotypes of 31 C-I accessions (50%) differed from each other only in the presence/ absence or the size of Giemsa C-bands in particular positions; this karyotypic variant was considered basic or "normal (N)." Karyotypes of 31 accessions derived from normal as a result of one or more structural chromosome rearrangements. Accessions collected from geographic closer regions usually had more similar banding patterns than accessions from distant locations, and this similarity remained in genotypes with chromosomal rearrangements. The highest C-banding and translocation polymorphisms were observed in Turkey. Ae. neglecta was similar to Ae. columnaris in the amount and distribution of Giemsabands on most chromosomes, but differed from it in the morphology of 6X c , which was more symmetric (arm ratio L/S = 1.173 vs. 1.924). In contrast to the C-II group, Ae. neglecta carried three pairs of the satellite chromosomes as the C-I accessions ( Fig. 1, a, d). Three accessions of Ae. neglecta had similar C-banding patterns (Fig. 3, t-v) and did not possess chromosomal rearrangements.
. 1U c -7X c -chromosomes; translocated chromosomes are indicated with arrows and designated respectively.
Lebanese group of Ae. columnaris contained eight accessions, one of which consisted of two karyotypically normal biotypes differing only in the C-banding patterns (Fig. 3, g, h). Of them, accession K-4241b (Fig. 3, h) was almost identical to K-4004 (Fig. 3, i) in the C-banding pattern. Most Lebanese Ae. columnaris had normal karyotypes, and two types of chromosomal rearrangements were identified in three accessions. Thus, K-4003 and K-4407 carried T7U c :2X c -T10 translocation (Fig. 3, j, k), while a pericentric inversion of the chromosome 6X c (inv2) was detected in K-4406 (Fig. 3, l).
Five accessions were from Syria. Three of them had normal karyotypes, and two different complex translocations were identified in the remaining two accessions (Fig. 3,  a, b). K-4372 carried T2U c :4X c + T4U c :6X c (T18) and K-4362 -the translocation T2U c :5X c + T4U c :2X c (T20). In both cases, the original single translocations were not found.
The origin of four accessions -AE 1512, AE1607, TX01, and CIae 34 was unknown. We found that AE 1607 consisted of two biotypes differing in chromosomal rearrangements (inv2X c / inv2X c -2 + T6U c :7X c ) and the C-banding patterns. This accession also contained admix of Ae. triuncialis seeds. Three accessions -CIae34, TX01, and AE 1521 carried pericentric inversion of 7U c (Fig. 3, p-r). This rearrangement was recorded only in Ae. columnaris collected from Central Anatolian in Turkey (Fig. 2, outlined with green dotted lines); therefore, we suggested that these three accessions may originate from the same region.
Seven Ae. columnaris accessions: six from Turkey and one from Iraq, were karyotypically distinct from all other accessions of the species and exhibited significant variation of the C-banding patterns (Fig. 7). They were assigned to the C-II group. Accession TA2084 carried at least two whole-arm reciprocal translocations; unidentifiable minor translocations may present in other accessions causing variation of the C-banding patterns. Despite heterogeneity, karyotypes of all C-II accessions shared some distinct features discriminating them from the C-I group and Ae. neglecta: (1) they had only two pairs of the satellite (SAT) chromosomes; (2) chromosome 1U c was more heterochromatic, whereas (3) chromosome 4U c of C-II contained less heterochromatin compared to C-I (Fig. 7).
(4) chromosome 7U c did not possess a prominent C-band complex in a proximal part of the long arm, found in the orthologous chromosomes of all C-I or Ae. neglecta accessions.
Morphology and the C-banding pattern of the chromosome 5U c in both C-I and C-II accessions were similar; however, 1U c of C-II was more heterochromatic than the 1U c in C-I (Figs. 7, 8). Significant differences existed in C-banding patterns of other C-I and C-II chromosomes, although some polymorphisms could result from introgression. Thus, chromosome 3U c of PI 554182 (Fig. 7, d) had the C-banding pattern typical for Turkish and Transcaucasian C-I accessions (i.e., PI 554186, PI 554187 on Fig.4, e, i) and may originate via introgression between C-I and C-II groups. A C-banding pattern of the chromosome 4X c of PI 564180 was more similar to 4X t of Ae. neglecta (Fig. 3, t-v) than other C-II or C-I accessions.

FISH analysis of Ae. columnaris
In order to get a deeper insight into genetic differences between groups C-I and C-II of Ae. columnaris and to assess their relationship with Ae. neglecta, we applied FISH with pTa71 (45S rDNA), pTa794 (5S rDNA), GAAn, GTTn, ACTn, pSc119.2, pAs1, pTa-713 probes for their analysis. The pTa-535 probe was not considered because it produced signals only on a few chromosomes (Fig. S1, o; green signals), uninformative for our analyses.
Hybridization of pTa71 and pTa794 probes revealed three pairs of major, nearly equal pTa71 signals on chromosomes of C-I and Ae. neglecta, but only two pairs of major NORs in the C-II accessions ( Fig. 8; Fig. S1, a, b; Fig. S2, a, c). Instead, all C-II accessions possessed faint pTa71 signals on a chromosome pair carrying a clear distal 5S rDNA locus. This chromosome was classified as 1X c based on results of sequential FISH with 5S + 45S rDNAs followed by GAAn + GTTn/pTa-713 probes (Fig. S2, b, d).
An additional minor NORs were found in the middle of 6U*L (Fig. S3) of all Ae. columnaris and Ae. neglecta. Ae. neglecta differed from Ae. columnaris in the presence of a minor 45S rDNA site in a distal part of an arm of a pair of large metacentric X*genome chromosome tentatively classified as 6X t (Fig. 8; Fig. S1, c, arrowed; Fig. 2, e; Fig. S3a). The application of FAM-labelled oligo-probes allowed us to detect very weak minor pTa71-signals at the terminus of 5X*L, a distal quarter of 1U*L, and in a proximal part of 3X c S (Fig. S3). Similar signals were obtained on chromosomes of the C-II accession PI 564181 (data not shown). However, these minor sites never appeared when the plasmid DNA was used as a probe, and they were not considered in the analysis. Apparent differences between C-I, C-II, and Ae. neglecta groups existed in a pattern of 5S rDNA probe. All Ae. columnaris, C-I and Ae. neglecta accessions contained ten 5S rDNA signals distributed among four chromosome pairs (Fig. S2, a, e; Fig. S3). The chromosome 1X * possessed two pTa794 sites: one located distally to the NOR, while the second -proximally to it ( Fig. 8; Fig. S1, a, c; Fig. S2, a, e). By contrast, four chromosome pairs in all C-II accessions carried a single 5S rDNA signal each.
In Ae. columnaris and Ae. neglecta labeling patterns of GAAn probe were mainly consistent with the C-banding patterns, while the GTTn hybridized predominantly on the X c chromosomes (Fig. S1, g-i). Only 2U * , 4U *, and 5U * contained small GTTn sites in pericentromeric/ proximal regions, and a faint signal was present in the middle of the 7U c L arm (Fig. 8) in four of the five C-II accessions. By contrast, all X c genome chromosomes demonstrated prominent GTTn signals located predominantly in the proximal chromosome regions. Position of the GTTn clusters on the X c chromosomes only partially overlapped with the GAAn location; some chromosomes (e.g., 5X c or 7X c ) were poorly labeled with GAAn, showed extremely heavy labeling with GTTn (Fig. 8).
Hybridization patterns of ACTn probes were almost identical to that of GTTn (Fig. S1, m,  n).
The pSc119.2 probe hybridized to subtelomeric regions of one or both arms of most Ae. columnaris chromosomes except for 7X c , which lacks pSc119.2 signals in all C-I and most C-II accessions ( Fig. 8; Fig. S1, j, k, o, n). Intercalary sites appeared only on the long arm of 7U c and rarely of 6U c L, as in a diploid Ae. umbellulata. Labeling patterns of the pSc119.2 probe were polymorphic between and among C-I and C-II accessions (Fig. 8). Four of the five C-II accessions studied by FISH possessed intercalary pSc119.2 site also on the chromosome 2U c L (Fig. 8, c), but this site was never observed in C-I or Ae. neglecta. On the other side, the C-II accession PI 564181 did not possess any pSc119.2 signals on the chromosome 2U c (Fig. 8, i).
The D-genome specific probes pAs1 and especially pTa-535 were low informative for chromosome identification of Ae. columnaris and Ae. neglecta. Distinct pAs1 sites were observed in the pericentromeric region of 6U * and 4X * chromosomes of all studied species, whereas 2-3 weak signals were present on their 4X * S and 7X * L arms. The chromosome 5X c of C-I also contained a single, small pAs1 site in the distal half of the short arm. Hybridization sites of the pTa-535 probe emerged on the 6U c L arm, but only in a few accessions studied (Fig. 8, m; Fig. S3, b).
The pTa-713 probe hybridized to most Ae. columnaris (C-I and C-II) and Ae. neglecta chromosomes, while the 3U c , 4X c, and 5X c (in Ae. neglecta -also 7X t ) lacked the signals completely. In most cases, the distribution of pTa-713 sites on chromosomes of all three groups was similar; however, some differences between them were observed ( Fig. 8; Fig. S4). In particular, a large pTa-713 signal was present on the short arm of 1U c of all C-I and Ae. neglecta accessions, but it was absent in the C-II. Most C-I and one Ae. neglecta (K-4233) possessed a distinct site in the proximal half of 2U c L, which was not found in C-II and two Ae. neglecta accessions (Fig. S4, g, h). We did not observe proximal pTa-713 sites in the short arm of 2X c L and 3X c L in the C-II group, but they were present in all C-I and Ae. neglecta accessions. In turn, a large pTa-713 signal present on the 1X c L arm of all C-II accessions was never observed in the C-I group or Ae. neglecta (Fig. S4). Position of hybridization sites on 5U c , 6U c , 7U c , was similar in all three groups, but Ae. columnaris differed from Ae. neglecta in morphology and/ or labeling patterns of 6X * and 7X * chromosomes (Figs. 8).

Analysis of gliadin spectra of Ae. columnaris
Electrophoretic analysis revealed a high diversity of gliadin spectra in 25 Ae. columnaris accessions and their distinctness from the spectra of Ae. neglecta (Fig. S5). Only two of 25 Ae. columnaris accessions, K-4413 and K-4418 from Iran, shared similar gliadin spectra, whereas four C-II accessions included in our analysis were highly diverse. However, all contained electrophoretic (EP) components, whose position did not match the overall pattern specific for Ae. columnaris C-I accessions (Fig. 9).
Thus, electrophoretic profiles of PI 564180 and PI 542191 were characterized by lowintense, virtually invisible ("minor") components in the α-zone; their intensities and position were distinct from other accessions of Ae. columnaris (Fig. 9 c, d; Fig. S5). Based on comparison with the K-1193 spectrum, we proposed that these components can be encoded by both the X c and U c genomes (Fig. 9).  [27] are shown schematically at the right side of the electrophoretic spectrum; (b) EP spectrum of the accession TA2084 in comparison with wheat cultivar Bezostaya-1; (c) EP spectrum of PI 564180; (d) EP spectra of Ae. columnaris accessions illustrating protein components presumably encoded by the X c (red dots) and U c (yellow dots) chromosomes. The unique components, which were not found in any other Ae. columnaris accessions, are shown schematically (parts b, c, and d).
Protein components located in the ω-zone of the spectra of all Ae. columnaris C-I accessions were similar in intensity and position (Fig. S5). Among them, components designated as "2" and "3" (Fig.9, indicated with red dots) corresponded to components detected in the spectrum of K-1193, which were coded by the 1X c chromosome. In contrast to other materials, accession PI 564181 contained the unique double band instead of "component 3" (Fig. 9d). Besides, it displayed a distinct distribution of components located in the β -γ zone, which, by comparison with the K-1193 spectrum, can be coded by group-6 chromosomes of the U c and X c genomes. Such distribution was more typical for common wheat, and the respective zone was controlled by wheat chromosomes 6B and 6D [356].
Protein components encoded by chromosome 1U c were characterized by low intensities (Fig. 9a, 9c; indicated by yellow dots). By contrast, the spectra of TA 2084 and PI 564180 possessed several intense components in the upper part of ω-zone designated 1', 2', and 3'. By analogy with the spectrum of K-1193, we hypothesized that they could be controlled by the chromosome 1U c (Fig. 9b, c). TA 2084 and PI 564180 spectra shared components 2' and 3'with similar mobility and intensity, but they differed in the presence of additional minor component 1', which showed slower mobility in TA2084.
Variability of the U-genome specific U31 nuclear fragment in Ae. columnaris and Ae. neglecta Amplification and further sequencing of the U-genome-specific U31 nuclear fragment was performed with primers U31a and U31b in fifteen accessions, including ten Ae. columnaris (K-4225, K-4228, K-4409, K-4413, and PI554186 from different countries and representing chromosomal group C-I, and PI542191, PI564179, PI 564180, PI564181, and TA2084 all from Turkey representing group C-II), two Ae. neglecta from Algeria and Turkey (PI 170209 and AE 646) in comparison with three accessions of their diploid parental species Ae. umbellulata (AE 1339, AE 155, and AE 820) of different geographic origin (Table 1). All accessions analyzed generated 363 bp fragments, except for Ae. columnaris PI 554186. In this accession, the fragment length was reduced to 270 bp due to a 123 bp deletion ( Fig. 10; Fig S6).
The sequence of the U31 fragment obtained from Ae. columnaris accessions fall into the types, which corresponded to designations proposed earlier by (Kadosumi et al. 2005) based on fragment length and the presence of MspI restriction site (CCGG). Type-I having the full-size U31 fragment and an intact MspI site was found in seven Ae. columnaris accessions as well as in all analyzed Ae. neglecta and Ae. umbellulata accessions ( Fig. S6; Fig. 10). Figure 10. Nucleotide substitutions in the U31 region in 15 Ae. columnaris (U c U c X c X c ), Ae. neglecta (U t U t X t X t ), and Ae umbellulata (UU) sequences. Dots correspond to nucleotides identical to consensus sequences. The MspI restriction site is highlighted in red.
The type-II U31 fragment was identified in two Ae. columnaris accessions, both from the C-II chromosomal group (Fig. 10). It emerged as a result of sequence changes at the MspI restriction site: a mononucleotide deletion in position 292 was found in TA2084, while C/T290 substitution in PI 564181. Accession PI 554186 (C-I) possessed the type-III U31 fragment with a 123 bp deletion (Fig. S6). All U31-alleles assigned to type-II corresponded to those reported by Kadosumi et al. (2005) in Ae. columnaris or Ae. neglecta. Among U31 type-I accessions of Ae. columnaris, four allelic variants were found, three of which were novel alleles (Fig. 10). Two of them were identified in C-II and one in C-I accession.
The U31 sequences of Ae. umbellulata accessions AE 155 and AE 820 of both Ae. neglecta (PI 170209 and AE 646) belonged to type-I and showed just a few (1-2) nucleotide substitutions, while almost 12 SNPs were detected in the U31 sequence of Ae. umbellulata, AE 1339 from Greece, which was also assigned to type-I (Fig. S6).
Most of the U31 alleles of Ae. umbellulata or Ae. neglecta discovered by us (Fig. 10) were not identified earlier, and only Ae. neglecta, PI 170209 carried the same allele as Ae. columnaris (KU-2953A) from Armenia, described earlier by Kadosumi et al. [23].
An ML tree (Fig. 11) shows the possible evolutionary relationship between accessions and species based on comparative sequencing of the U31 alleles. All Aegilops accessions except AE 1339 (Ae. umbellulata) formed one common cluster on the tree obtained. No species-specific or ploidy-specific clusters have been observed. Three Ae. columnaris accessions including two of type-II U31 alleles (PI 564179, PI 564180) and one type-III accession (PI 554186) formed a separate sub-cluster with 79% bootstrap support. Other accessions representing different species (Ae. columnaris and Ae. neglecta) and different U31 allele types (I and II) fall into one common sub-cluster with Ae. umbellulata (AE 115 and AE 820) showing a closer relationship. Aegilops neglecta accession (AE 646) form an individual branch. Figure 11. Maximum-likelihood (Kimura 2-parameter model) tree of the U-genomespecific U31 nuclear sequence. The numbers above the branches indicate bootstrap values; the C-banding group is shown in red, U31 allele type -in green.
According to the analysis of all three plastome regions, ten Ae. columnaris accessions split into two groups (plastogroups). Four C-I accessions (K-4225, K-4228, K-4409, and PI 554186) had identical sequences of the plastome spacers, while K-4413 (C-I, Iran) differed at a single site: substitution of the hexamer sequence CCTCAT by ATGAGG at position 470-475 of the rpl32-trnL spacer (Fig. 12). Accessions of group C-II: PI542191, PI564179, PI 564180, PI564181, and TA2084, showed significantly higher sequence polymorphisms at all three plastome regions. Nevertheless, they all shared the same deletion of one of the two AAGAA 5-bp repeats, as well as the deletion of the mononucleotide T at position 446. (Fig. 12). Besides, they all carried G/T substitution at position 468 of trnT-trnL, as Ae neglecta and Ae. umbellulata accessions.
C-II accession PI 564179 possessed the highest number of mutations, especially in the rpL32-trnL sequence. Together with C-II accession PI 564181 and Iranian C-I K-4413, it carried ATGAGG/CCTCAT sequence substitution. The same substitution was also identified in Ae. umbellulata and Ae. neglecta (Fig. 12). Comparison of the observed plastogroups with groups discriminated based on C-banding and FISH analyses showed that all Ae. columnaris accessions characterized by an increased variability (PI542191, PI564179, PI 564180, PI564181, and TA2084) belonged to karyotypic group C-II, while low polymorphic accessions (K-4225, K-4228, K-4409, K-4413, PI 554186) fall in C-I. On ML tree (Fig. S7), all Ae. columnaris accessions with invariable plastome sequences clustered together, whereas K-4413 formed a separate branch in a common sub-cluster with two Ae. umbellulata accessions (bootstrap = 67). Five genetically variable Ae. columnaris accessions fall either in a common sub-group with Ae. neglecta (TA2084, PI 542191, PI 564180), or formed separate branches (PI 564179, PI 564181) (Fig. S7).

Discussion
Cytogenetic (C-banding, FISH), biochemical (seed storage proteins -gliadins), and molecular (sequence analysis of polymorphic U31 nuclear fragment and three intergenic regions of cpDNA) analyses showed close genetic relationships of Ae. columnaris and Ae. neglecta, in agreements with previous studies [2,4,11,12,15,19,23]. From another side, chromosome analysis revealed higher genetic diversity of Ae. columnaris compared to that reported for Ae. neglecta [4,13,23], which was expressed in higher C-banding/ FISH-polymorphisms and broader spectra of chromosomal rearrangements as well as by a higher number of U31 alleles and higher variability of cpDNA identified in these species.
Two karyotypic groups, the C-I and C-II, have been discriminated within Ae. columnaris based on chromosome analysis and each group displayed characteristic C-banding and FISH patterns, enabling their discrimination. Group C-I was mainly similar to Ae. neglecta, whereas C-II differed from the C-I group of Ae. columnaris and from Ae. neglecta in karyotype structure, heterochromatin distribution, and in the patterns of rDNA loci. Such heterogeneity of ribosomal loci was not reported for other Aegilops species [13,[36][37][38][39]. Although these karyotypic groups were not supported by comparing gliadin profiles or sequences of the U31 nuclear fragment, they fully coincided with plastogroups discriminated based on cpDNA analysis.
Groups C-I and C-II karyotypically differed from each other, but the divergence level varied between individual chromosomes. Thus, no significant changes were observed in 2U c , 5U c , 2X c , and 6X c , while 3U c , 4U c , 7U c , 1X c , 5X c , 7X c of the C-II were modified. Despite it, we found chromosomes among C-II accessions, which matched chromosomes of C-I (e.g., 3U c of PI 554182) or Ae. neglecta (e.g., 4X c of PI 564180), which can be caused by introgressions. Another evidence of gene flow between species and chromosomal groups came from the analysis of the U31 nuclear fragment: Type-II U31 allele identified in C-II accession PI 564181 ( Fig. 12) was earlier detected by Kadosumi et al. [23] in four accessions of Ae. neglecta and three Ae. umbellulata, but not in Ae. columnaris. A similar trend was observed in the presence of ATGAGG/CCTCAT substitution in the rpl32-trnL spacer region, which was present in one C-I and two C-II accessions of Ae. columnaris, but also in Ae. neglecta and Ae. columnaris (Fig. 12).
All methods used in our study highlighted significant genetic diversity of both C-I and C-II chromosomal groups, but each of them exhibited a different type of polymorphisms. Karyotype divergence in the C-I group was associated with variation in the presence and size of C-bands in particular positions and chromosomal rearrangements identified here in 55% of the accessions studied. However, no polymorphisms that could be associated with introgressions or unbalanced rearrangements have been found. The results of electrophoretic analysis of seed storage proteins led to the same conclusion. Although 25 accessions of Ae. columnaris had unique gliadin profiles; the spectra of most C-I genotypes shared several characteristic bands, especially in the -zone. The number of U31 alleles identified here in the C-I accessions (Fig. 10) was relatively small, and this group displayed very low polymorphism of the intergenic spacers of cpDNA: only one 6-bp-substitution in position 470 of rpl32-trnL was found (Fig. 12).
By contrast, accessions constituting the C-II group were highly heterogeneous. Although karyotypes of all accessions carried several diagnostic features discriminating them from the C-I group and Ae. neglecta, the observed variation cannot be explained by polymorphism of heterochromatic regions only. The emergence of some variants can be due to introgressions and heterochromatin re-pattering. In contrast to group C-I, chromosomal rearrangements did not play such an essential role in the divergence of the C-II group: translocations were detected only in TA2084, which is geographically distant from others (Table 1; Fig. 2). However, minor translocations may present in other C-II accessions, but they skip identification due to the lack of appropriate markers. Significant heterogeneity of the C-II group was also shown by gliadin analysis. All four C-II accessions had different gliadin profiles, which did not possess any common components. The spectra of each of the C-II accessions (PI 564180, PI 564181, TA2084, and PI 542191), however, carried a number of features (band loss or gain; bands that differed in intensity or position) which were not observed in the C-I group.
The comparative sequence analysis of the U31 nuclear fragment and three plastome intergenic spacer regions also revealed the highly heterogeneous composition of the C-II groups. Thus, the U31 fragment of type-I was found in three C-II accessions (Fig. 10), but two of them carried mutant alleles. All accessions with type-II U31 fragment belonged to the C-II group. It was an interesting observation because, according to Kadosumi et al. [23], type-II U31 fragment occurred extremely rare in Ae. columnaris, although frequently in Ae. umbellulata. Both type-II alleles identified here in the C-II accessions corresponded to those described earlier by these authors, but they found one allele in Ae. columnaris from Syria, while the second -in Ae. neglecta. Kadosumi et al. [23] also identified an additional type-II U31 allele, not found in this work, in Ae. columnaris from Iran; however, the karyotypic group of this accession was not determined.
In contrast to the relatively conservative C-I group, from three to 27 SNP's covering all three intergenic spacer regions of cpDNA were identified among accessions of the C-II group.
An interesting fact uncovered by molecular analysis of the U31 nuclear fragment was an unexpectedly high number of SNPs (12) identified in Ae. umbellulata accession AE 1339 from Greece (Fig. S6), which showed no changes in the cpDNA (Fig. 12). According to FISH [40], this accession was karyotypically normal and similar to other Ae. umbellulata genotypes in the distribution of repetitive DNA probes [4,[41][42][43][44][45][46]. All these indicate that the observed mutations of AE 1339 were not caused by chromosomal rearrangement. From another side, Kawahara [47] has already uncovered the distinctness of Ae. umbellulata population from Greek Islands based on morphological and isozyme markers.
Summarizing our results, we can conclude that Ae. columnaris is phylogenetically very close to Ae. neglecta, and probably derived from it (or their common ancestor). It is supported by the following observations.
1. Owing to species-specific inversion of the chromosome 6X c , the karyotype of Ae. columnaris becomes more "asymmetric" compared to Ae. neglecta. According to Stebbins [48], an increase of karyotype asymmetry is a trend of evolution in plant species and, therefore, Ae. neglecta karyotype should be considered "more primitive," while Ae. columnaris -"more advanced;" 2. The chromosome 6X t of Ae. neglecta possesses a minor 45S rDNA locus, which probably pre-existed in the progenitor Aegilops species; however, this locus is absent in Ae. columnaris;

3.
Ae. columnaris is characterized by chromosome instability expressed in a higher proportion and broader diversity of chromosomal rearrangements (20 variants in more than 55% of accessions). Chromosome instability is an essential factor of speciation [49,50] and is usually more expressed in phylogenetically new species. In addition, we found significant intraspecific polymorphism of Ae. columnaris plastome, although the only low variation of the chloroplast DNA sequences was recorded in Triticum and Aegilops species [51,52].
The similarity of rDNA and repetitive DNA patterns of chromosomes of Ae. neglecta and group C-I of Ae. columnaris and their distinctness from chromosomes of the C-II accessions indicate that the C-I group diverged from Ae. neglecta or their common ancestor as a result of minor genome modifications. Group C-II could derive from a progenitor presumably belonging to group C-I of Ae. columnaris relatively recently, probably due to introgression from another Aegilops species, accompanied by significant reorganization of the parental genomes. As most C-II accessions with known collection sites originated from the very narrow geographic region of the southeastern coastal part of Turkey (Fig. 2, red boxes), they might originate from one common ancestor. Significant heterogeneity of the C-II accessions in karyotype structure, Cbanding and FISH patterns, gliadin composition, and nuclear and chloroplast DNA sequences may indicate that they are currently at the initial stage of species divergence; most likely, this group is undergoing an extensive speciation process.

Materials and Methods
Intraspecific diversity of Aegilops columnaris Zhuk. (2n=4x=28, U c U c X c X c ) was assessed on a set of 69 accessions of various geographic origin in comparison with the related tetraploid species Ae. neglecta Req. ex Bertol. (2n=4x=28, U t U t X t X t ) -4 accessions and Ae. umbellulata Zhuk. (2n=2x=14, UU), the diploid U-genome progenitor of Ae. columnaris and Ae. neglecta -3 accessions. All 69 accessions were analyzed using C-banding, while 16 Ae. columnaris, three Ae. neglecta, and two Ae. umbellulata accessions were studied by FISH (Fluorescence in situ hybridization). Gliadin profiles were examined in 25 Ae. columnaris accessions of various geographic origins and one Ae. neglecta (Table 1), whereas 10 Ae. columnaris (five from C-I and five from C-II groups), two Ae. neglecta and three Ae. umbellulata accessions were selected for subsequent molecular analysis.

Giemsa C-banding method
The C-banding procedure was carried out as described in Badaeva et al. [60]. Chromosomes of Ae. columnaris were classified according to genetic nomenclature developed earlier by [20] based on analysis of introgressive lines. Chromosomes of Ae. neglecta were classified according to similarity with Ae. columnaris chromosomes. Designation of Ae. umbellulata chromosomes followed the nomenclature suggested by [41].

Fluorescence in situ hybridization
FISH was carried out according to the protocol described in [61]. The probes labeled with fluorescein were detected using anti-fluorescein/Oregon green®, rabbit IgG fraction, Alexa Fluor® 488 conjugate (Molecular Probes, USA). The slides were counter-stained with DAPI (4',6-diamidino-2-phenylindole) in Vectashield mounting media (Vector Laboratories, Peterborough, UK) and examined on a Zeiss Imager D-1 microscope. Selected metaphase cells were captured with AxioCam MRm digital camera using software AxioVision, version 4.6. Images were processed in Adobe Photoshop R , version CS5 (Adobe Systems, Edinburgh, UK).

Seed storage protein (Gliadin) analysis
Electrophoresis (EP) in polyacrylamide gel (PAAG) according to the previously published protocol [62] was employed to obtain gliadin spectra of the 25 Ae. columnaris and one Ae. neglecta accessions. The spectra of wheat cultivar Bezostaya-1 (a standard of gliadin spectra of common wheat) and Ae. columnaris, K-1193 with the known genetic control of gliadin components [27] were used to compare gliadin profiles of other Aegilops accessions (Fig. 2, a).
DNA extraction, PCR amplification, and DNA sequencing Ten accessions of Ae. columnaris (five C-I representing five countries and five C-II from Turkey), Ae. umbellulata (3 accessions) and Ae. neglecta (2 accessions) were selected for analyses by molecular methods. Genomic DNA was extracted from 10-day-old seedlings using the DNeasy Plant Mini kit (QIAGEN, Hilden, Germany). DNA quantitative and qualitative evaluation was performed using NanoDrop 2000c spectrophotometer (Thermo-Scientific, Madison, USA). Amplification of the U-genome-specific U31 nuclear fragment was performed using primers U31a and U31b [23] with PCR reaction conditions: an initial denaturation step of 95 0 C for 5 min followed by 30 cycles of 94 0 C for 1 min, 55 0 C -1 min, and 72 0 C for 1 min with a final extension step at 72 0 C for 3 min. The amplified fragments were sequenced directly from both ends with the same U31a and U31b primers. Amplification of the three intergenic spacers regions (trnH(ugu)-psbA, rpl32-trnL(tag), trnT(ugu)-trnL(uaa)) of the plastome DNA of Aegilops accessions was performed using primer sets listed in Table S1. PCR amplification was performed in a 15-μl reaction mixture containing approximately 50 ng genomic DNA, 1.5 μl of 10× PCR buffer, 1.5 mM MgCl2, 0.2 mM of dNTPs, 0.3 μM of each primer, and 0.5 unit of Taq DNA polymerase. The PCR conditions were as follows: an initial denaturation step of 95 0 C for 5 min, followed by 30 cycles of 94 0 C for 1 min, annealing at the appropriate Tm for 1 min, and 72 0 C for 1 min with a final extension step at 72 0 C for 5 min. Amplification temperatures for trnH-psbA was 58 0 C, trnL-rpl32 -56 0 C, trnT-trnL-55 0 C. The same primers were used to sequencing the obtained chloroplast DNA fragments; PCR products were cleaned before sequencing using the QIAquick PCR purification kit (QIAGEN, Hilden, Germany). PCR products were sequenced using standard protocols with the ABI Prism Big Dye Terminator cycle sequencing kit v. 3.1. Sequences were resolved on an ABI Prism 3100 automated sequencer. A phylogenetic tree was constructed based on U31 data and combined chloroplast sequence data using MEGA 7 [63] based on ML (maximum likelihood) method. Kimura 2-parameter model was used for U31 and Tamura-3 parameter model for cpDNA, which was selected using Modeltest; 1000 bootstrap replicates were applied for the branch support evaluation. The SNP data from 10 Ae. columnaris genotypes were taken for subsequent analyses. The SNP position was determined from the first nucleotide of U31 or of each of the analyzed chloroplast spacers.

Funding
The work was performed within the framework of the State Task according to the