We present the most comprehensive survey of cytogenetic and genomic diversity of GGAtAt wheats. We describe the composition, distribution and characteristics of the GGAtAt genepool. Building on our results, the latest published complementary genomics studies and state-of-the-art archaeobotanical evidence we revisit the domestication history of the GGAtAt wheats. We arrived at the following four key findings:
1. The GGAtAtgenepool consists of three distinct lineages
We sampled the full breadth and depth of GGAtAt wheat diversity and discovered a clear genetic and geographic differentiation among extant GGAtAt wheats. Surprisingly, and supported by all marker types, three clearly distinct lineages were identified. The first lineage is comprised of all T. timopheevii genotypes (and the derived T. militinae and T. zhukovskyi; note that all T. militinae and all T. zhukovskyi accessions maintained ex situ in genebanks are each derived from only one original genotype). Interestingly, T. araraticum consists of two lineages that we preliminarily describe as ‘ARA-0’ and ‘ARA-1’. This finding is in contrast to Kimber and Feldman (1987) who concluded that T. araraticum does not contain cryptic species, molecularly distinct from those currently recognized. Based on passport data, ARA-0 was found across the whole predicted area of species distribution. ARA-1 was only detected in south-eastern Turkey and in neighboring north-western Syria. It is interesting to note that only in this part of the Fertile Crescent, the two wild tetraploid wheat species, T. dicoccoides and T. araraticum, grow in abundance in mixed stands.
Based on Fig. 3, collecting gaps are evident for T. araraticum, and future collecting missions should focus on four specific regions: (i) between the Euphrates river in the west and the Elazığ – Silvan – Mardin transect region in the east. Interestingly, so far only T. dicoccoides was reported from this region; (ii) between Bitlis (Turkey) – Amadiyah (Iraq) in the west and the Armenian border in the east including north-western Iran; (iii) between Adıyaman – Silvan in the south and Tunceli in the north; and (iv) between Hama in the south and west/north-west of Aleppo in Syria.
We found only two T. araraticum populations which contained representatives of both ARA lineages: (i) 45 km south-east of Kahramanmaraş to Gaziantep, and (ii) 4 km north of St. Simeon on the road to Afrin in Syria. However, in both cases, the ARA-1 lineage was significantly more frequent than ARA-0. This suggests at least a certain level of taxon boundary between ARA-0 and ARA1 lineages and should be investigated in the future.
Independent support for the existence of two wild T. araraticum lineages and their distribution comes from Mori et al. (2009) based on 13 polymorphic chloroplast microsatellite markers (cpSSR) (Supplementary Table S2, column 15). The ‘plastogroup G-2’ was distributed in south-eastern Turkey and northern Syria and was closely related to Triticum timopheevii (Mori et al. 2009). However, Gornicki et al. (2014), based on whole chloroplast genome sequence information and sufficient taxon sampling (13 Triticeae species and 1127 accessions; 163 accessions in common with our study, Supplementary Table S2, column 17), provided increased resolution of the chloroplast genome phylogeny and showed that the T. timopheevii lineage possibly originated in northern Iraq (and thus according to our data, belong to the ARA-0 lineage as no ARA-1 occurs in Iraq). This was supported by Bernhardt et al. (2017), who, based on re-sequencing 194 individuals at the chloroplast locus ndhF (2232 bp) and on whole genome chloroplast sequences of 183 individuals representing 15 Triticeae genera, showed that some ARA-0 and TIM genotypes are most closely related. All GGAtAt wheats re-sequenced by Bernhardt et al. (2017) were considered in our study (Supplementary Table S2, column 16). Haplotype analysis of the Brittle rachis 1 (BTR1-A) gene in a set of 32 T. araraticum in comparison with two T. timopheevii accessions (Nave et al. 2021) also showed closer relationships of domesticated T. timopheevii to wild T. araraticum from Iraq. That is more, one of these accessions, TA102 (PI 538461, 1 km NE of Salahaddin) shared the same haplotype with T. timopheevii and it was assigned to ARA-0 group by our study (Supplementary Table S2, column 18).
It is important to note that our results (i.e., the characteristics, composition and geographical distribution of ARA-0 and ARA-1 lineages) are not in agreement with the latest comprehensive taxonomical classification of wheat by Dorofeev et al. (1979), who divided T. araraticum into two subspecies: subsp. kurdistanicum Dorof. et Migusch. and subsp. araraticum (Supplementary Table S2, column 2). We propose to re-classify the GGAtAt genepool taxonomically in the future.
2. The karyotypic composition of GGAtAtwheats is as complex as the phylogenetic history of the GGAtAtgenepool
Based on karyotype analyses, translocation spectra and distribution of DNA probes, T. araraticum populations from Dahuk and Sulaymaniyah (both Iraq) harbored the highest karyotypic diversity among all T. araraticum populations studied. We consider the region around Dahuk in Northern Iraq as the center of diversity of T. araraticum, and this is probably the region where T. araraticum originated. This is supported by Nave et al. (2021), who found the highest haplotype diversity among T. araraticum from Iraq, and by Bernhardt et al. (2017) and Gornicki et al. (2014) who traced chloroplast haplotypes from Aegilops speltoides growing in Iraq via T. araraticum (ARA-0) to T. timopheevii and T. zhukovskyi.
The karyotype ‘similar’ to those in ‘normal’ T. timopheevii was found in 44.6% of all T. araraticum genotypes. This is the group of candidates, in which the closest wild relative(s) to T. timopheevii is (are) expected. The frequency of the normal karyotype varied among countries and between populations (Supplementary Figure S12). It is interesting to note that the Samaxi-Akhsu population in Azerbaijan and some populations near Dahuk (Iraq) possessed mostly karyotypically normal genotypes. Diagnostic C-bands for the ARA-1 lineage, both with normal and rearranged karyotypes, were 1AtL3, 4AtS7, 5AtL3, 1GL5, 2GL7, 3GL7 + L11 (Fig. 1). As expected, the number of C-bands characteristic for ARA-0 was smaller (due to the wide geographical distribution) and only two C-bands were lineage-specific and found in normal as well as translocated genotypes: 6AtL3 and 5GL15.
However, some FISH patterns suggested that T. timopheevii probably originated in Turkey and probably from ARA-1 (or, ARA-1 and TIM may have originated from a common ancestor, but then diverged). This is supported by the following observations: (i) TIM and ARA-1 carry the pSc119.2 signal in the middle of 1At long arm, while this site was absent from ARA-0; (ii) all ARA-0 and most ARA-1 possessed the Spelt-52 signal on 6GL, but it is absent in all TIM and five ARA-1 genotypes from Gaziantep-Kilis, Turkey. The distribution of Spelt-1 and Spelt-52 probes on chromosomes of these five genotypes was similar to, and in accession IG 116165 (ARA-1 from Gaziantep) almost identical with TIM; (iii) the pAesp_SAT86 patterns on chromosomes 3At, 4G, and 7G are similar in TIM and ARA-1 but differed from ARA-0. Differences between ARA-1 and TIM based on FISH patterns of some other chromosomes as well as the results of C-banding and molecular analyses suggest that extant ARA-1 genotypes are not the direct progenitors of TIM but that the ARA-1 lineage is most closely related to it.
Based on AFLP, C-banding, FISH and Jeli retrotransposon markers, TIM was genetically most closely related to ARA-1. Additional evidence for the close relationship between TIM and ARA-1 lineages comes from allelic variation at the VRN-1 locus of genome At (Shcherban et al. 2016). This analysis revealed a 2.7 kb deletion in intron 1 of VRN-A1 in three T. timopheevii and four T. araraticum accessions, which, according to our data, belong to the ARA-1 lineage. However, at Vrn-G1, TIM from Kastamonu in Turkey (PI 119442) shared the same haplotype (Vrn1Ga) with ARA-1 samples, while TIM from Georgia harbored haplotype VRN-G1 as found in ARA-0. These results suggest multiple introgression events and incomplete lineage sorting as suggested by Bernhardt et al. (2017, 2020).
Regular chromosome pairing observed in the F1 hybrids of lines with ‘normal’ karyotypes (Kawahara et al. 1996), identified in our study as ARA-0 × ARA-1 (Supplementary Table S2, column 14), suggested that karyotypic differences between ARA-0 and ARA-1 lineages are not associated with structural chromosomal rearrangements such as large translocations or inversions.
The emergence or loss of most lineage-specific Giemsa C-bands (Fig. 3) or FISH loci (Supplementary Fig. 22, Supplementary Fig. 23) could be due to heterochromatin re-pattering: amplification, elimination or transposition of repetitive DNA sequences. Wide hybridization can also induce changes in C-banding and FISH patterns of T. araraticum chromosomes. Changes in pAesp_SAT86 hybridization patterns on 4G and 7G chromosomes, however, are likely to be caused by pericentric inversions, which are also frequent in common wheat (Qi et al. 2006). The role of inversions in inter- and intraspecific divergence is probably underestimated. In our case, it seems possible that divergence between ARA-1/TIM (two inversions) from ARA-0 (no inversion) was associated with at least two pericentric inversions.
We did not find any genotype harboring both ARA-0 and ARA-1 specific FISH sites, although ARA-0 and ARA-1 genotypes co-existed in two populations in Turkey and Syria. However, based on FISH (Spelt-1 site on chromosome 6GL and 7GS, respectively), hybridization between certain ARA-1 and ARA-0 lines can be predicted.
Iran occupies a marginal part of the distribution range of T. araraticum. An abundance of the pericentric inversion of the 7At chromosome in the Iranian group indicates that it is derived from Iraq. The karyotypically ‘normal’ genotype was probably introduced to Transcaucasia via Western Azerbaijan (Iran). The low diversity of FISH patterns and the low C-banding polymorphism of T. araraticum from Transcaucasia indicate that T. araraticum was introduced as a single event. Interestingly, the AFLP data suggested some similarity between ARA-0 from Armenia and Azerbaijan and T. timopheevii.
We hypothesize that homoploid hybrid speciation (HHS) (Abbott et al. 2010; Nieto Feliner et al. 2017; Soltis and Soltis 2009) and incomplete lineage sorting may be the possible mechanisms explaining the origin of the ARA-1 lineage. Although this assumption was not experimentally supported, it is favored by some indirect evidence. ARA-1 grows in sympatry and in mixed populations with T. dicoccoides (Fig. 3) and is phylogenetically most closely related to T. dicoccoides (Supplementary Table S6; Supplementary Table S8). ARA-1 is morphologically more similar to T. dicoccoides. Thus, five of 10 misclassified T. araraticum accessions belonged to ARA-1 group (Supplementary Table S2), two of which, PI 656871 and IG 116176, were the mix of T. dicoccoides and ARA-1. Five misclassified ARA-0 accessions from USDA-ARS collection were from Siirt, Turkey, however in other gene bank three of these accessions were treated as T. araraticum. Relatively good chromosome pairing was observed in the F1 hybrids of some T. araraticum x T. timopheevii combinations (Tanaka and Ichikawa 1972), however, pollen fertility of such hybrids was very low (0.3–5.4%). ARA-1 could be derived from ancient hybridization of T. timopheevii × T. dicoccoides; or alternatively, ARA-1 and TIM could be derived from the hybridization ARA-0 × T. dicoccoides.
3. Does theT. timopheeviipopulation found in western Georgia represent the last remnant of a widespread ancient cultivation area of GGAtAtwheats?
Wild emmer T. dicoccoides belongs to the first cereals to be domesticated by humans in the Fertile Crescent and the evolution and domestication history of T. dicoccoides are relatively well studied (Badaeva et al. 2015b; Civáň et al. 2013; Özkan et al. 2011). Domesticated emmer Triticum dicoccon Schrank was a staple crop of Neolithic agriculture, was widely cultivated for over 10,000 years and harbored impressive genetic diversity (Nesbitt and Samuel 1996; Szabo and Hammer 1996; Zaharieva et al. 2010). The domestication of T. dicoccoides provided the key for durum wheat (Maccaferri et al. 2019) and bread wheat evolution (Pont et al. 2019).
Much less is known about the domestication history of T. timopheevii. It is believed that T. timopheevii is the domesticated form of T. araraticum (Dorofeev et al. 1979; Jakubziner 1932). In contrast to T. dicoccon, T. timopheevii was, since its discovery, considered as a ‘monomorphous narrowly endemic species’ (Dekaprelevich and Menabde 1932) cultivated in few villages of western Georgia (Stoletova 1924-25; Zhukovsky 1928) (Supplementary Material S27). Dekaprelevich and Menabde (1932) noticed that the area of cultivation had probably been larger in the past. The last plants of T. timopheevii in situ were found by the expedition of the N.I. Vavilov Institute of Plant Genetic Resources (VIR, Russia) in 1983 near the village of Mekvena (Tskhaltubo, Georgia) and deposited in the VIR genebank under accession number K-56422 [E.V. Zuev, personal communication]. Today, the widespread view is that the cultivation area of T. timopheevii was restricted to Georgia in the (recent) past (Feldman 2001; Mitrofanova et al. 2016, Zohary et al. 2012).
However, hulled tetraploid wheat morphologically similar to T. timopheevii was identified at three Neolithic sites and one Bronze Age site in northern Greece and described by Jones et al. (Jones et al. 2000) as a ‘New’ Glume Wheat (new glume wheat, NGW). The glume bases of these archaeological finds morphologically resemble T. timopheevii more than any other extant domesticated wheat (Jones et al. 2000). After these finds of NGW in Greece, this wheat was also identified at Neolithic and Bronze Age sites in Turkey, Bulgaria, Romania, Hungary, Slovakia, Austria, Italy, Poland, Germany and France (Bieniek 2002, 2007; Bogaard et al. 2007, 2013; Ergun 2018; Fairbairn et al. 2002; Fiorentino and Ulaş 2010; Fischer and Rösch 2004; Hajnalová 2007; Kenéz et al. 2015; Kohler-Schneider 2003; Kreuz and Boenke 2002; Perego 2017; Rottoli and Pessina 2007; Toulemonde et al. 2015; Ulaş and Fiorentino 2020; Valamoti and Kotsakis 2007). Earlier finds of an ‘unusual’ glume wheat in Serbia (Borojevic 1991) and Turkey (de Moulins 1997) have subsequently been recognized as NGW (Kenéz et al. 2015; Jones et al. 2000; Kroll 2016). Criteria were also established for distinguishing the grains of NGW (Kohler-Schneider 2003).
At some sites, NGW appeared as a minor component and may have been part of the accompanying weed flora of cereal fields (Kenéz et al. 2014; Ulaş and Fiorentino 2020). In other cases, it was probably cultivated in a mix with einkorn (Jones et al. 2000; Kohler-Schneider 2003) and/or emmer. The recovery of large quantities in storage deposits of whole spikelets at Çatalhöyük in Turkey, caryopses and spikelet bases at the early Bronze Age settlement of Clermont-Ferrand in France, and rich deposits including whole spikelets at bronze age sites in Italy, demonstrated that, at least in some places, NGW was a major crop in itself (Bogaard et al. 2013, 2017; Ergun 2018; Kenéz et al. 2014; Perego 2017; Toulemonde et al. 2015).
Based on intensive archaeobotanical investigations at Çatalhöyük in central Anatolia, for example, NGW was the predominant hulled wheat, overtaking emmer wheat around 6500 cal BC and remaining dominant until the site’s abandonment c. 5500 cal BC. The finds suggested that this wheat was a distinct crop, processed, stored, and presumably grown, separately from other glume wheats. NGW formed part of a diverse plant food assemblage at Neolithic Çatalhöyük, including six cereals, five pulses and a range of fruits, nuts and other plants, which enabled this early farming community to persist for 1500 years (c. 7100 to 5500 cal BC) (Bogaard et al. 2013, 2017).
Recently, polymerase chain reactions specific for the wheat B and G genomes, and extraction procedures optimized for retrieval of DNA fragments from heat-damaged charred material, have been used to identify archaeological finds of NGW (Czajkowska et al. 2020). DNA sequences from the G genome were detected in two of these samples, the first comprising grain from the mid 7th millennium BC at Çatalhöyük in Turkey, and the second made up of glume bases from the later 5th millennium BC site of Miechowice 4 in Poland. These results provide evidence that NGW is indeed a cultivated member of the GGAtAt genepool (Czajkowska et al. 2020). As NGW is a recognized wheat type across a broad geographic area in prehistory, dating back to the 9th millennium BC in SW Asia, this indicates that T. timopheevii (sensu lato, s.l. = domesticated GGAtAtwheat in general), was domesticated from T. araraticum during early agriculture, and was widely cultivated in the prehistoric past (Czajkowska et al. 2020).
This raises the question of whether the few populations of T. timopheevii (sensu stricto, s.str.) found in western Georgia were the last remnants of a wider GGAtAt wheat cultivation or whether the T. timopheevii of Georgia was a local domestication independent of the domestication of T. araraticum in SW Asia. To answer this question, sequence information for NGW, the Georgian T. timopheevii, and the two lineages of T. araraticum (ARA-0, ARA-1) need to be compared.
Vavilov (1935) suggested that T. timopheevii of western Georgia was probably originally introduced from north-eastern Turkey. As cited by Dorofeev et al. (1979), Menabde and Ericzjan (1942) associated the origin of T. timopheevii with the region of the ancient kingdom of Urartu, whence immigrant ancestors of modern-day Georgians introduced it into western Georgia. Certainly, the possibility of introduction of Timofeev’s wheat into Georgia from the south should not be rejected (Dorofeev et al. 1979). Is the domestication history of T. timopheevii s.str. connected with other endemic wheats of Georgia, such as T. karamyschevii Nevski (T. georgicum Dekapr. et Menabde or T. paleocolchicum Menabde) and T. carthlicum Nevski, which were cultivated by Mingrelians in Western Georgia (Jorjadze et al. 2014; Mosulishvili et al. 2017)?
4. Was the cultivation range of T. timopheevii (s.str.) wider in the recent past?
We screened all available passport data and found three cases, which could potentially help to reconstruct the recent past cultivation range of T. timopheevii s.str: Interestingly, two T. timopheevii accessions maintained in two ex situ genebanks are reported to originate from Turkey (Supplementary Table S2) [https://www.genesys-pgr.org/]: (i) ATRI 3433 (TRI 3433) conserved in the Federal ex situ Genebank of Germany hosted at the Institute of Plant Genetics and Crop Plant Research, IPK, Gatersleben. This T. timopheevii line was most likely collected by E. Baur in Turkey in 1926 (Schiemann 1934); and (ii) PI 119442 identified among a barley sample obtained from a market in Araç, near Kastamonu, Turkey (Fig. 3) in 1936 by Westover and Wellmann, and maintained at the National Plant Germplasm System, USDA-ARS, USA. Both accessions harbor the ‘normal’ karyotype of T. timopheevii, both were characterized in our studies and confirmed as a typical T. timopheevii. Additionally, the accession TA1900, presumably collected 32 km south of Denizli near Kahramanmaraş in Turkey on the 14th of August 1959 and maintained in the wheat germplasm collection of the Wheat Genetics Resource Center, Kansas State University, U.S.A., is interesting because it shared karyotypic features of T. timopheevii and the ARA-0 lineage (Fig. 4; Supplementary Figure S19, i31). However, we are not fully convinced that this line is a true natural hybrid. Based on the passport data, this accession could potentially have escaped from an experimental field or a breeding station, or received introgression(s) during ex situ maintenance (Zencirci et al. 2018). Assuming that the passport data is correct, we could speculate that T. timopheevii may have been cultivated in Turkey during the first half of the 20th century. However, is this realistic option?
We believe not. As reported by Stoletova (1924-25), Dekaprelevich and Menabde (1929, 1932), Menabde (1948), Dekaprelevich (1954), T. timopheevii s.str. was part of the spring landrace Zanduri (mixture of T. timopheevii s.str. and T. monococcum) and well adapted to the historical provinces Lechkhumi and Racha of Georgia (Supplementary Material S28). The Zanduri landrace was cultivated in the ‘humid and moderately cool climate zone 400–800’ m above the sea level (Dorofeev et al. 1979). Martynov et al. (2018) reported that T. timopheevii potentially has a ‘low potential for plasticity’ and is not drought tolerant. Climate at origin based on bioclimatic variables (Fick et al. 2017; R Core Team 2017) clearly differs between the regions of Western Georgia where the Zanduri landrace grew till the recent past, and both Kastamonu and Kahramanmaraş regions in Turkey (Supplementary Material S28). We speculate that the three T. timopheevii accessions which were collected in Turkey were probably introduced from Transcaucasia or elsewhere and may have been left over from unsuccessful cultivation or breeding experiments of T. timopheevii s.str. in the recent historical past. Also, based on botanical records, T. timopheevii (s.str.) has only been identified in Georgia, but not in Turkey or elsewhere (Davis 1965–1988; Hanelt 2001). From this we conclude that the cultivation range of T. timopheevii (s str.) was not wider in the recent past.