Seeds of wild emmer accession PI 428082 were received from the US National Small Grains Collection, Aberdeen, Idaho. The accession was collected 52.2 km west of Diyarbakir (Latitude: 37° 46 min 59 s, Longitude: 39° 46 min 0 s) in the foothills (elevation 1400 m) of Karaca Dağ. The accession was crossed as the male parent with Langdon, which was received from L.R. Joppa (University of North Dakota, Fargo). Five F₁ plants were self-pollinated to produce an F₂ generation. The single seed descent technique was followed to advance generations by self-pollination in a greenhouse. A total of 445 independent F₆–F₈ RILs were developed.

DNAs were isolated from leaf segments (Dvorak et al., 2006b) and genotyped using the wheat 90K iSelect Infinium genotyping assay (Illumina Incorporated, San Diego, CA, United States) at the UC Davis Genotyping Core facility. Output was analyzed with the GenomeStudio program (Illumina, San Diego, CA, United States). Genotype data were uploaded to the Mutipoint Ultra-dense (ULD) mapping program (MultiQTL Limited, Haifa, Israel) and processed as a RIL population. Only grouped (co-segregating) markers (Ronin et al., 2017) were employed in the first round of linkage map construction (marker “clustering”). This resulted in 55 linkage groups (LGs). Markers within a LG were ordered relative to each other using a “hard” setting of marker order monotony control. Marker order monotony was subsequently visually inspected, and markers that disturbed monotony were removed. In the next step, the ends of each LG were extended with singleton markers. The extended LGs that were within 0.1 recombination frequency (RF) were merged. Finally, singleton markers that showed linkage to internal marker groups in a LG were inserted.

The LGs were exported from Multipoint ULD to Microsoft Excel (Microsoft, Corp., Seattle, WA, United States) and compared to the 90K consensus map. This comparison was used to assign LGs to chromosomes, to determine their orientation, and to detect chimeric LGs. Observed RF values were then transformed into per-meiosis values (Haldane and Waddington, 1931) to compensate for the accumulation of recombination events across generations, from F₁ to F₆–F₈. Lastly, centimorgan (cM) distances were computed from the transformed RF values using the Kosambi mapping function (Kosambi, 1943). The genetic map was compiled in Microsoft Excel, and a genotype matrix following the marker order on the genetic map was created. Graphical genotypes (Young and Tanksley, 1989) were scrutinized using a custom script using Python (Python Software Foundation, Beaverton, OR, United States) for inconsistent data points within the occasional heterozygous segments. The inconsistent data points were removed as part of quality control and all data points within a heterozygous block were labeled as missing data because the Multipoint ULD program could not be run in the specified mode with heterozygous data. This revised matrix was then used as the final input for the Multipoint ULD program to construct the final genetic map.

The nucleotide sequences for the wheat 90K Illumina Infinium SNP markers were downloaded from the database (Wang et al., 2014) and used as BLASTN 2.2.28+ (NCBI) queries to search for homologous sequences in the A. tauschii pseudomolecules (GenBank BioProject PRJNA341983). A database was created, in which each SNP marker was associated with an A. tauschii subject hit. Colinearity between linkage map SNP markers and the A. tauschii pseudomolecules was assessed by searching for an ascending or descending order of top hit (subject) locations on the A. tauschii pseudomolecules. The arbitrary condition for declaring a marker colinear was that it was a member of a group of at least three different loci (genes) in a colinear order; otherwise markers were considered non-colinear.

Groups of colinear markers that indicated a chromosome rearrangement relative to A. tauschii, such as an inversion or translocation, were compared with consensus genetic maps of durum and common wheat (Maccaferri et al., 2014; Wang et al., 2014). Additional comparisons were made with the Brachypodium distachyon, v3.1 (Initiative International Brachypodium Genome, 2010), rice, v7.0 (Matsumoto et al., 2005), and sorghum v3.1 (Paterson et al., 2009) pseudomolecules to validate each structural change and to determine its ancestral versus derived state.

Segregating markers were used to compute recombination rates, expressed as cM/Mb, using the cM position of a marker in the LG. Since a wheat genome sequence needed for the computation of recombination rates was not available to us we used the A. tauschii pseudomolecules (GenBank BioProject PRJNA341983) as the most closely related reference for the A and B genomes of tetraploid wheat. To estimate these rates, we employed local cubic kernel derivative smoothers with Gaussian kernel using the package KernSmooth in R (The R Foundation, Vienna Austria). The bandwidth used was 20 Mb and was chosen manually (Fan and Gijbels, 1996).

A table (Supplementary Table S4 in Akhunov et al., 2010) containing SNP diversity statistics for expressed sequence tag (EST)-derived sequences in 10 accessions of wild emmer from the Karaca Dağ region and 13 accessions of T. aestivum ssp. aestivum, compactum, and spelta, was downloaded. The starting nucleotides on the A. tauschii pseudomolecules (GenBank BioProject PRJNA341983) for ESTs for which diversity data existed in the table were determined and the locations of the ESTs on the genetic map were imputed using the locations of the 90K wheat iSelect Infinium markers on the A. tauschii pseudomolecules as references. Diversity statistics, Watterson nucleotide polymorphism estimator θw (Watterson, 1975), nucleotide diversity θπ (Nei and Li, 1979), and Tajima’s D (Tajima, 1989) were computed (Akhunov et al., 2010) and θπ was graphed to assess diversity distribution along the genetic map of chromosome 4A in Karaca Dağ wild emmer and T. aestivum.

DNAs from the 445 independent F₆ to F₈ RILs from the cross Landon x PI 428082 were genotyped with the 90K wheat SNP iSelect Infinium assay. Sixteen (3.7%) RILs were removed from the population because of various genotyping defects, leaving 429 RILs for further analyses. The 90K Infinium assay contained 26,385 D-genome markers and 55,038 A- and B-genome markers. Only the latter were a priori relevant to tetraploid wheat genotyping. Of these, 13,422 (24.1%) markers were polymorphic between Langdon and PI 428082 and generated well-clustered genotyping graphs with GenomeStudio. A small portion, 138 (0.5%), of the markers classified as D-genome markers in the 90K Infinium database (Wang et al., 2014) also generated well-clustered SNP genotyping graphs.

In total, 13,560 markers produced well-clustered genotype data and were used in map construction. After two cycles and manual editing of data for spurious genotype calls, a map consisting of 10,914 markers was produced (Table 1). The map was comprised of 2,650 segregating markers (referred in Multipoint ULD program as skeleton markers), which were single markers representative of a bin of co-segregating markers (Supplementary Table S1) and 8,264 co-segregating markers (referred in Multipoint ULD program as bound markers) (Supplementary Table S2).

Fourteen LGs were obtained. Twelve were consistent with the wheat 90K Infinium consensus map (Wang et al., 2014), one was chimeric and consisted of arms 3BL, 6BS, and 6BL, and one contained arm 3BS. The latter LG merged with the chimeric LG when the requirement RF < 0.1 was relaxed. The linkage between proximal markers in 3BS and 6BL and 6BS and 3BL indicated that the chimeric LG was caused by a 3B-6B reciprocal translocation in wild emmer with breakpoints in the 3B and 6B centromeric regions. The four chromosome arms making up the 3B-6B reciprocal translocation were purposefully kept as separate LGs throughout the work reported here even though 3BS and 6BL LGs and 6BS and 3BL LGs were linked across the centromeres. The resulting 16 LGs (Supplementary Figure S1) had a total length of 1,727.93 cM and had an average of one segregating marker every 0.65 cM. The lengths of the LGs of chromosomes that were not involved in the 3B-6B translocation ranged from 101.55 cM for 6A to 176.48 cM for 5A (Table 1).

For downstream applications, it was desirable to use all markers present on the map, not just the segregating markers. To order the 8,264 co-segregating (bound) markers that co-segregated with the segregating (skeleton) markers, the sequences of all 10,914 markers were used as queries in BLASTN homology searches against the A. tauschii pseudomolecules (GenBank BioProject PRJNA341983) and hits with expect value < E-5 were recorded (Supplementary Table S3). The segregating markers were then arranged according to their locations in the LGs whereas co-segregating markers in each co-segregating bin were ordered according to their starting nucleotides on the A. tauschii pseudomolecules so that their ascending or descending progression was consistent with the progression of the neighboring segregating markers (Supplementary Table S3). Of the 10,914 markers, 9,131 (83.6%) were ordered using this strategy (Table 1 and Supplementary Table S3).

Some markers hit many sites on the A. tauschii pseudomolecules in BLASTN searches (Supplementary Table S4), suggesting that they may have been derived from repeated sequences. The highest number of hits was 7,165 for marker Tdurum_contig28050_299. We chose > 10 hits in the A. tauschii pseudomolecules as an arbitrary threshold for considering a SNP marker to be derived from a repeated sequence. Using this threshold, 42 (0.4%) of the 10,914 SNP markers were derived from repeated sequences (Supplementary Table S4).

This expanded genetic map and marker locations on the A. tauschii pseudomolecules were used to compute recombination rates along the 14 chromosomes, which were expressed as cM per Mb (Supplementary Figure S2).

Disregarding temporarily the previously known structural rearrangements involving chromosomes 4A, 5A, and 7B, 15 structural rearrangements relative to the order of markers along the A. tauschii pseudomolecules were found (Table 2). Recombination was detected within all rearrangements suggesting that, except for the 3B-6B translocation, all were likely shared by the parents and were homozygous in the F₁ generation.

The most frequent type of chromosome rearrangement was an intrachromasomal translocation. Nine were detected and all were short, ranging in length from 134,271 to 6,145,185 bp as measured on A. tauschii pseudomolecules, and in all of them a chromosome segment was translocated only a short distance (Table 2). Our map shared most of the intrachromosomal translocations with the durum consensus map (Maccaferri et al., 2015) or common wheat consensus map (Wang et al., 2014) (Table 2).

The second most frequent type of chromosome rearrangement was a paracentric inversion. Three were large, Inv(1) detected in arm 3AS, Inv(2) detected in arm 7AL, and Inv(3) juxtaposed in arm 7AL to Inv(2) (Table 2). The fourth inversion, Inv(4), was only 0.23 cM long, and was at the tip of arm 7BS, in the 5AL segment translocated to 7BS. Our map shared the order of markers in these four inversions with the durum consensus map (Table 2), and in Inv(1), also with that on the consensus map of common wheat (Table 2).

There were two interchromosomal translocations. One was between 6BS and 7BL. The segment present in 7BL was absent on the 6BS genetic map, suggesting that the rearrangement was non-reciprocal. The translocation was also present on the durum consensus map (Maccaferri et al., 2015) (Table 2). The other interchromosomal translocation was the reciprocal translocation 3B-6B described above. No segmental duplication satisfying our arbitrary requirement to involve three consecutive genes was detected on the genetic map.

The marker progressions within the inverted and flanking regions in LGs 3A, 3B, 7A, 7B, and the homoeologous A. tauschii, B. distachyon, rice, and sorghum pseudomolecules were determined (Supplementary Table S5) to distinguish between the ancestral and derived states of marker order for these specific rearrangements. For Inv(1), LGs 3A and 3B and the Bd2, Os1, and Sb3 pseudomolecules had the same marker order whereas pseudomolecule 3D had the alternative order (Supplementary Table S5). The fact that the outgroup B. distachyon, rice, and sorghum pseudomolecules had the same marker order as LG3A and 3B indicated that this order was the ancestral state and that of 3D was a derived (inverted) state.

No markers were mapped on the 7B map in the region corresponding to Inv(2) and the proximal part of Inv(3). Therefore, only a distal portion of Inv(3) was studied in colinearity comparison including homoeologous regions of LG7A and 7B and pseudomolecules 7D, Bd1, Os6, and Sb10. Marker order was shared by LG7A and Bd1, Os6, and Sb10 whereas LG7B and the 7D pseudomolecule shared the alternative order (Supplementary Table S5). Following the same rationale as above, we concluded that the 7A marker order was the ancestral state and that in LG7B and pseudomolecule 7D was the derived (inverted) state.

The alignment of the chromosome 4A, 5A, and 7B markers on the A. tauschii pseudomolecules confirmed the locations and orientations of major segments previously reported in these rearranged chromosomes (Devos et al., 1995; Mickelson-Young et al., 1995; Nelson et al., 1995; Miftahudin et al., 2004; Hernandez et al., 2012). The following complex sequence of events leading to the evolution of the rearranged chromosomes 4A-5A-7B has been proposed (Figure 1) (Devos et al., 1995; Mickelson-Young et al., 1995; Miftahudin et al., 2004): (1) A reciprocal 4AL-5AL translocation, which exchanged the distal portions of 4AL and 5AL arms. This translocation originated in the diploid ancestor of the wheat A genome. (2) A pericentric inversion in 4A, which converted the short arm to the present-day long arm and a remnant of the long arm became the present-day short arm. (3) A reciprocal translocation between 4AL (originally 5AL) and 7BS. (4) A paracentric inversion that inverted the 4AL and 5AL segments in the 4AL arm. (5) A small 4A pericentric inversion. Our genetic map in concert with the A. tauschii reference sequence confirmed rearrangements (1), (3), and (5) but revealed the following discrepancies.

We obtained an unequivocal evidence for the presence of genes of the ancient 4AL at the end of the ancient 4AS (present-day 4AL) but could not confirm the presence of two EST markers (BE518074 and BE494743) in the terminal region of the present-day arm 4AS (Supplementary Table S6). No other short arm marker from this region was mapped distal to the ancient 4AL fragment making up the present-day 4AS (Supplementary Table S3). We therefore failed to obtain evidence for the second breakpoint of the hypothetical large pericentric inversion.

The breakpoints of the distal paracentric inversion (4) in the present-day 4AL were proposed to be in the short 4AL segment and the 7BS segment (Figure 1). If that were true, a portion of the 7BS segment should have been moved to a proximal position relative to the 5AL segment and a portion of the 4AL segment should have remained proximal to the 5AL segment, as shown in Figure 1. Proximal locations relative to the 5AL segment of two 4AL ESTs (BE499664 and BE637934) and four 7BS ESTs (Supplementary Table S6) were used as supporting evidence for these breakpoints (Miftahudin et al., 2004). However, none of the six ESTs were proximal to the 5AL segment on our genetic map and genes homologous to them were in different locations in the A. tauschii reference genome sequence (Supplementary Table S6). The ancient 4AL, 5AL, and 7BS segments making up the distal portion of the present-day 4AL were intact on our genetic map and showed no signs of breakpoints causing the paracentric inversion (4).

The short arm of the present-day chromosome 4A comprises only 76% of the proximal portion of the ancient 4AL chromosome arm. Terminal deletions of this size greatly reduce or entirely preclude meiotic pairing with the wild-type homologs (Curtis et al., 1991). The long arm of the rearranged wheat chromosome 4A is composed of four segments. Of these, only the terminal 7BS segment shares a telomere and orientation with the ancient chromosome 7B and could have paired and recombined with it (Figure 1). The remaining segments are in different positions and inverted orientations compared to the ancient wild-type chromosomes (Figure 1). As a result, the rearranged wheat chromosome 4A could not have paired with the wild-type chromosomes, except for the distal 7BS segment, and except for this segment, the rearrangements would have severely limited gene flow into the rearranged chromosome 4A.

Selection for the rearranged chromosome 4A would have led to a selective sweep involving most of the chromosome, except for the terminal 7BS segment. That being the case, most if not all diversity that exists in the rearranged chromosome 4A, except for the 7BS segment, must have originated since the evolution of the present-day chromosome 4A. Hence, nucleotide diversity of wheat 4A relative to that in the rest of the A-genome can be used to assess the relative age of the rearranged wheat chromosome 4A. The more recent are the rearrangements, the less diverse the chromosome will be relative to the remaining six A-genome chromosomes.

Akhunov et al. (2010) reported nucleotide diversity at 2,114 expressed sequence loci based on Sanger sequencing in 12 accessions representative of T. aestivum and 10 accessions of wild emmer from the Karaca Dağ region (Figure 2). The data included 110 4A loci. Nucleotide diversity in 4A was significantly lower than the A-genome population mean (Akhunov et al., 2010) but without a genetic map, the distribution of genetic diversity along the rearranged chromosome 4A could not have been fully interpreted (Akhunov et al., 2010).

We used our genetic map of 4A in re-analyzing these data by re-computing the mean Watterson nucleotide polymorphism measure θw, nucleotide diversity θπ, and Tajima’s D for six A-genome chromosomes without 4A and for 4A itself (Table 3). In T. aestivum, θw but not θπ was significantly lower in 4A than in the remaining six A-genome chromosomes (Table 3). In wild emmer from the Karaca Dağ region, both diversity measures were significantly lower in 4A than in the remaining A-genome chromosomes (Table 3).

In T. aestivum, θπ was uniformly high along 4A (Figure 3). For an unknown reason, loci with diversity estimates reported by Akhunov and his colleagues were disproportionally under-represented in the long arm of 4A, particularly in the 5AL, 4AL, and 7BS segments (Figure 3). Nevertheless, in the few loci that were investigated in these segments nucleotide diversity was similar to that in the rest of the chromosome.

In Karaca Dağ wild emmer, the centromeric region involving 30 loci showed very low diversity indicating a selective sweep (Figure 3 and Supplementary Table S7). While θπ was 0.51 × 10^-3 and 0.96 × 10^-3 in the distal regions of wild emmer 4A (Table 3), θπ was only 0.01 × 10^-3 in the centromeric region, although θπ was 0.65 × 10^-3 in the same region in T. aestivum. In the sweep area, Tajima’s D was highly negative in wild emmer (Table 3), which is indicative of recent selective sweep followed by population expansion.

Of the 90K Infinium SNP markers 13,560 were polymorphic between Langdon and wild emmer accession PI 428082 and 2,650 were mapped with the population of 429 RILs as segregating (skeleton) markers. For comparison, 16,387 90K Infinium markers were polymorphic in a mapping population of 150 F₆ RILs from a cross between durum ‘Svevo’ and wild emmer ‘Zavitan’ from northern Israel (Avni et al., 2014) but only 2,297 were mapped as segregating (skeleton) markers (Avni et al., 2014). This difference in mapping outcomes highlights the importance of the number of RILs in the mapping population for mapping efficiency.

Recombination rates were high in the distal regions of chromosomes and declined toward the proximal regions. This pattern is consistent with other recombination rate studies in wheat and its close relatives in the tribe Triticeae (Dvorak and Chen, 1984; Lukaszewski and Curtis, 1993; Dubcovsky et al., 1996; Gill et al., 1996; Zhang et al., 2001; Akhunov et al., 2003; Luo et al., 2005, 2013; Avni et al., 2014). In the Langdon x PI 428082 population, the recombination rates declined more precipitously in the short arms than in the long arms. In the long arms of chromosomes 1, 4, and 5 and the arms of large metacentric chromosomes 2, 3, and 7, the rates peaked about 50 Mb from chromosome termini (Supplementary Figure S2).

An important factor affecting recombination in wheat is the homoeologous pairing suppressor Ph1. The suppressor acts on polymorphism between recombining chromosomes (Luo et al., 1996). The greater is polymorphism, the greater is recombination rate suppression. This inverse relationship accounts for the short lengths of LGs observed on our map, particularly in the B genome chromosomes which, for an unknown reason, are affected by polymorphism more than the A-genome chromosomes (Dvorak and McGuire, 1981). While the average A-genome LG was 130.3 cM, the average B-genome LG was only 116.4 cM (P = 0.03, paired t-test).

A factor that undoubtedly confounded estimation of recombination rate in the B genome was heterozygosity for the 3B-6B reciprocal translocation. Reciprocal translocations are common in wild emmer and are more frequent in the B-genome than in the A-genome (Kawahara, 1986, 1987). We do not know the frequency of the 3B-6B translocation described here in the wild emmer population, because accession PI 428082 was not included in the Kawahara’s study. Heterozygosity for a reciprocal translocation reduces recombination rates in the chromosome arm that includes a break (Dobzhansky, 1931). The total length of chromosome 3B (sum of the 3BS and 3BL LGs) was 116.2 cM and total length of chromosome 6B (sum of the 6BS and 6BL LGs) was 95.8 cM. Both chromosomes were shorter than the mean genetic length of the remaining five B-genome chromosomes, 120.6 cM. The short length of the two chromosomes was particularly notable for chromosome 3B which is physically the largest wheat chromosome (Dvorak et al., 1984). Both 3B and 6B were genetically also shorter than their A-genome homoeologs.

The comparison of the genetic map with the A. tauschii reference genome sequence uncovered 15 chromosome rearrangements, in addition to the known 4A-5A-7B rearrangements. Four of these were paracentric inversions, Inv(1) to Inv(4). Since inversion heterozygosity suppresses recombination in the inverted region (except for two-strand double crossovers), high levels of recombination within the inverted regions in the Langdon x PI 428082 F₁ indicate that the F₁ plants were homozygous for these inversions and that these inversions are shared by Langdon and PI 428082. Moreover, all four inversions were present on the durum consensus map (Maccaferri et al., 2015) suggesting that they are widely distributed in wild and domesticated tetraploid wheat.

Comparisons of the A-genome and B-genome LGs with the A. tauschii, B. distachyon, rice, and sorghum pseudomolecules showed for Inv(1) that the derived (inverted) state is found in 3D. A similar analysis showed for Inv(3) that the derived (inverted) state is shared by 7B and 7D. The ancestral vs. derived state of Inv(2) could not be investigated. Because the likelihood of a reversion of an inverted segment is small, sharing of inversions among genomes can be used to reconstruct phylogeny, as shown for the A, B, and D genomes (Figure 4).

Nine of the rearrangements were intrachromosomal translocations. In all of them, the translocated segment was short and was translocated only a short distance along the chromosome. It is tempting to attribute these translocations to transposition, since transposable elements (TEs) tend to transpose short distances. However, the same outcome could be produced by intrachromosomal crossover between TEs, which excises a circular intermediate that can be reinserted in the vicinity. It is therefore impossible to say without more analytical work whether the intrachromosomal translocations originated via transposition or ectopic recombination.

Several attempts based on RFLP and deletion maps have been made to reconstruct the evolution of the rearranged wheat chromosome 4A (Devos et al., 1995; Mickelson-Young et al., 1995; Miftahudin et al., 2004). Our wild emmer genetic map failed to confirm the current models of evolution of this chromosome. Specifically, it failed to validate one breakpoint of the pericentric inversion (2) and both breakpoints of the paracentric inversion (4) previously reported based on EST locations on the 4A maps. We failed to find the 7BS EST loci proximal to the 5AL segment and confirmed thus a similar failure to locate these EST loci in the survey sequence of 4A (Hernandez et al., 2012). A radiation hybrid map of 4A revealed discrepancies in the 4AL deletion breakpoints (Balcárková et al., 2017), which may account for the conflicting interpretation of EST locations. Clearly, the structure of chromosome 4A needs reassessment but we prefer the revisit it after a reference-quality sequence of wheat is available.

There are two sources of genetic diversity (θπ) in a polyploid species: introgression from the diploid progenitors (θπ_i) and mutations that have occurred since the origin of the polyploid species (θπ_p). Total diversity θπ of a polyploid is therefore θπ_i + θπ_p. The suppression of recombination and fixation of the rearranged chromosome 4A in wild emmer swept away θπ_i from the majority of 4A genes, and θπ_i can be assumed to be zero. The magnitude of θπ_p in 4A and θπ_p in the remaining A-genome chromosomes can therefore be used to estimate the age of the rearranged 4A chromosome relative to the age of wild emmer, provided that θπ_i in the remaining chromosomes can be estimated and subtracted from total diversity θπ. Wheat D genome is less than 8,000 years old, and most of its diversity, θπ = 0.18 × 10^-3, was contributed by gene flow from A. tauschii (Akhunov et al., 2010; Wang et al., 2013). We will therefore use the diversity in the wheat D genome as an estimate of θπ_i. Subtracting 0.18 × 10^-3 from the total diversity in the T. aestivum non-4A A-genome chromosomes estimates θπ_p = 0.42 × 10^-3. Remarkably, this estimate is identical to the estimate of total diversity θπ = 0.42 × 10^-3 in the 4A of T. aestivum (Table 3). The same manipulation estimates θπ_p = 0.55 × 10^-3 (0.73 - 0.18 × 10^-3) in the wild emmer non-4A A-genome chromosomes. This estimate is close to total diversity θπ = 0.50 × 10^-3 in wild emmer 4A.

We can take into consideration the diversity sweep in the wild emmer 4A and exclude that from the comparison. Then θπ_p = 0.51 × 10^-3 in the short arm of 4A and 0.96 × 10^-3 in the long arm of 4A (Table 3), making the agreement between total diversity in 4A and θπ_p in the remaining six A-genome chromosomes slightly closer.

It might be of interest to compare 4A diversity with diversity in the 5AL and 7BS segments involved in the 4A-5A-7B translocation. Unfortunately, diversity of only two genes in the 5AL segment and one gene in the 7BS segment was reported (Akhunov et al., 2010), which is inadequate for a meaningful comparison.

Both in T. aestivum and wild emmer, the levels of total diversity θπ in 4A are similar to θπ_p in the remaining six A-genome chromosomes. Assuming that diversity has been generated with equal rates in all A-genome chromosomes, these diversity levels suggest that the fixation of the rearranged chromosome 4A and the origin of wild emmer may have been contemporary or the rearrangements took place very early in the evolution of wild emmer.

The Karaca Dağ region includes several archeological sites on the upper Euphrates and Tigris rivers with some of the oldest records of agriculture, and this area is viewed by some as the cradle of agriculture in western Asia (Lev-Yadun et al., 2000). Evidence for domestication of emmer in the Karaca Dağ region (Nelson et al., 1995; Ozkan et al., 2002, 2005, 2011; Luo et al., 2007) is critical for this hypothesis.

If emmer were indeed domesticated in the Karaca Dağ region, wild emmer in the Karaca Dağ region would be the ancestor of all domesticated tetraploid and hexaploid wheat. Yet, evidence for gene flow between domesticated emmer and wild emmer in all areas where the two have been sympatric (Luo et al., 2007) raises concerns about the purity of wild emmer in general and in Karaca Dağ region in particular.

Another complicating factor is the selective sweep apparent in the centromeric region of chromosome 4A of wild emmer in the Karaca Dağ region. This sweep was previously detected with 10 RFLP loci in the centromeric region of 4A in a sample of 48 accessions from the Karaca Dağ region and 117 accessions of the northern population of domesticated emmer (Dvorak et al., 2006a). This sweep is perplexing in light of the fact that the same chromosome region in T. aestivum and wild emmer in other geographic regions (Avni et al., 2017) show normal levels of nucleotide diversity. If wild emmer in the Karaca Dağ region were ancestral to all domesticated wheat, what was the source of the diversity in T. aestivum? In addition, what is the cause of the selective sweep in the Karaca Dağ wild emmer population?

Diversity of the Psr920-4A RFLP locus (= ABCT-1 gene) suggests a possible answer to the first question. The locus is in the centromeric region of 4A and is dimorphic. With few rare exceptions, wild emmer from all areas of the Fertile Crescent has the Psr920b allele whereas all domesticated tetraploid wheats have the Psr920a allele (Dvorak et al., 2006a). Hexaploid wheat has both alleles (Dvorak et al., 2006a) suggesting introgression from wild emmer into chromosome 4A of T. aestivum. Thus, gene flow from wild emmer in regions outside Karaca Dağ could have possibly contributed diversity in the sweep area in T. aestivum.

We can provide no satisfactory explanation for the second question. A number of causes are possible, such as selection for adaptation to the Karaca Dağ environment or selection favoring a wild allele at a domestication gene on 4A. Since a similar selective sweep appears to exist also in the northern population of domesticated emmer, we cannot rule out even the extreme scenario that wild emmer in the Karaca Dağ region is actually feral (Civan et al., 2013) and thus derived from the northern population of domesticated emmer. These possibilities call for renewed attention to the purity of the Karaca Dağ wild emmer population and to the genetic relationships between it and other populations of wild and domesticated tetraploid wheat. The population of RILs developed here may be instrumental for mapping domestication genes in Karaca Dağ wild emmer and shedding light on this dilemma.