Genetic diversity, distribution and domestication history of the neglected GGAtAt genepool of wheat

Key message We present a comprehensive survey of cytogenetic and genomic diversity of the GGAtAt genepool of wheat, thereby unlocking these plant genetic resources for wheat improvement. Abstract Wheat yields are stagnating around the world and new sources of genes for resistance or tolerances to abiotic traits are required. In this context, the tetraploid wheat wild relatives are among the key candidates for wheat improvement. Despite its potential huge value for wheat breeding, the tetraploid GGAtAt genepool is largely neglected. Understanding the population structure, native distribution range, intraspecific variation of the entire tetraploid GGAtAt genepool and its domestication history would further its use for wheat improvement. The paper provides the first comprehensive survey of genomic and cytogenetic diversity sampling the full breadth and depth of the tetraploid GGAtAt genepool. According to the results obtained, the extant GGAtAt genepool consists of three distinct lineages. We provide detailed insights into the cytogenetic composition of GGAtAt wheats, revealed group- and population-specific markers and show that chromosomal rearrangements play an important role in intraspecific diversity of T. araraticum. The origin and domestication history of the GGAtAt lineages is discussed in the context of state-of-the-art archaeobotanical finds. We shed new light on the complex evolutionary history of the GGAtAt wheat genepool and provide the basis for an increased use of the GGAtAt wheat genepool for wheat improvement. The findings have implications for our understanding of the origins of agriculture in southwest Asia. Supplementary Information The online version contains supplementary material available at 10.1007/s00122-021-03912-0.


Introduction
The domestication of plants since the Neolithic Age resulted in the crops that feed the world today. However, successive rounds of selection during the history of domestication led to a reduction in genetic diversity, which now limits the ability of the crops to further evolve (Tanksley and McCouch 1997;van Heerwaarden et al. 2010). This is exacerbated by the demand for high crop productivity under climate change. Crop wild relatives (CWR) represent a large pool of beneficial allelic variation and are urgently required to improve the elite genepools (Dempewolf et al. 2017;Kilian et al. 2021). Bread wheat (T. aestivum L., 2n = 6x = 42, BBAADD) and durum wheat (T. durum Desf., 2n = 4x = 42, BBAA) are the staple crops for about 40% of the world's population. But as wheat yields are stagnating around the world (Iizumi et al. 2017;Ray et al. 2013Ray et al. , 2012, new sources of genes for resistance or tolerances to abiotic traits such as drought and heat are required. In this context, the wheat wild relatives are among the key sources for bread wheat and durum wheat improvement (Dante et al. 2013;Placido et al. 2013;Zhang et al. 2017).
Wild T. araraticum was first collected by M.G. Tumanyan and A.G. Araratyan during 1925-28 southeast of Erevan, Armenia (Tumanyan 1930;Nazarova 2007), soon after the discovery of domesticated T. timopheevii by P.M. Zhukovsky (Zhukovsky 1928) (Supplementary Material S1). Subsequently, T. araraticum was found in several other locations in Armenia and Azerbaijan (Dorofeev et al. 1979;Jakubziner 1933Jakubziner , 1959, as well as in Iran, Iraq and Turkey. Single herbarium specimens resembling T. araraticum have been sporadically recorded among T. dicoccoides accessions collected from the Fertile Crescent (Jakubziner 1932;Sachs 1953). However, only botanical expeditions from the University of California at Riverside (USA) to Turkey in 1965, to the Fertile Crescent in 1972-1973(Johnson and Hall 1967Johnson and Waines 1977) and the Botanical Expedition of Kyoto University to the Northern Highlands of Mesopotamia in 1970 (Tanaka and Ishii 1973;Tanaka and Kawahara 1976) significantly expanded our understanding of the natural distribution of T. araraticum. More recently, T. araraticum was found in northwestern Syria (Valkoun et al. 1998). Especially in southeastern Anatolia, Turkey, the distribution area of T. araraticum overlaps with the distribution range of T. dicoccoides. From the western to eastern Fertile Crescent, it is assumed that T. araraticum gradually substitutes T. dicoccoides (Johnson 1975), and T. dicoccoides is absent from Transcaucasia (Özkan et al. 2011). In most habitats, T. araraticum grows in patches and in mixed stands with other wild cereals (Troitzky 1932;Tumanyan 1930).
It is difficult to distinguish T. araraticum from T. dicoccoides by morphology under field conditions (Dagan and Zohary 1970;Tanaka and Sakamoto 1979). However, both species can easily be differentiated based on biochemical, immunological, cytological and molecular markers (Badaeva et al. 1994;Gill and Chen 1987;Jiang and Gill 1994;Kawahara and Tanaka 1977;Konarev et al. 1976;Lilienfeld and Kihara 1934). From an archaeobotanical perspective, both species can be reliably identified based on several characteristics of charred spikelets (Jones et al. 2000). In blind tests, it was possible to distinguish modern representatives of the two species with a c. 90% accuracy, on the basis of the primary keel of the glume, which arises just below the rachis disarticulation scar, and the prominent vein on the secondary keel (observable at the base of the glume, which is the part of spikelet most commonly preserved by charring in archeological material) (Jones et al. 2000).
Triticum dicoccoides is considered to be the older species than T. araraticum (Gornicki et al. 2014;Huang et al. 2002), which is supported by higher similarity of the S-G genomes compared to the S-B genomes (Jiang and Gill 1993;Kilian et al. 2007;Rodríguez et al. 2000a). Cytogenetic and molecular data showed that the speciation of T. araraticum was accompanied by complex species-specific translocations involving chromosomes 1G-6A t -4G and 3A t -4A t (Chen and Gill 1984;Jiang and Gill 1993;Rodríguez et al. 2000b;Salina et al. 2006) as well as with mutations of the primary DNA structure causing the divergence of homoeologous chromosomes (e.g., chromosomes 3A-3A t ) (Dobrovolskaya et al. 2009). These changes of karyotype structure are specific for the whole section Timopheevii as compared to the emmer wheat lineage (BBAA, BBAADD) (Badaeva et al. 1986;Hutchinson et al. 1982;Zhang et al. 2013).
Most recent phylogenetic studies based on whole chloroplast genome sequences, genome-wide sequence information and enlarged taxon sampling provided increased resolution of the evolutionary history within the Triticeae tribe, thereby shedding also new light on the GGA t A t wheat genepool (Bernhardt et al. 2017;Gornicki et al. 2014).
Research and pre-breeding activities have focused on T. dicoccoides because it gave rise to the economically most important wheats, T. durum Desf. and T. aestivum L. (Avni et al. 2017;El Haddad et al. 2021). However, T. araraticum and T. timopheevii have also contributed to bread wheat improvement. Several important genes controlling resistance against stem rust, leaf rust, powdery mildew or wheat leaf blotch (Allard and Shands 1954;Brown-Guedira et al. 1996, 2003Dyck 1992;McIntosh and Gyarfas 1971) were transferred to common wheat from T. timopheevii. Cytoplasmic male sterility (CMS) induced by T. timopheevii cytoplasm showed great potential for heterotic hybrid technology (Maan and Lucken 1972;Mikó et al. 2011;Würschum et al. 2017). However, despite its potential huge value for bread and durum wheat improvement, only a comparatively small number of genes were transferred from T. timopheevii (even less from T. araraticum). Most of the gene contributions originated from only one line (D-357-1) bred by R. Allard at the University of Wisconsin in 1948 (Martynov et al. 2018).
Understanding the population structure and the intraspecific variation of the entire tetraploid GGA t A t genepool would further its use for wheat improvement. In this study, the first comprehensive survey of cytogenetic and genomic diversity sampling the full breadth and depth of the tetraploid GGA t A t genepool is reported. We provide new insights into the genetic relationships among GGA t A t wheats and its domestication, taking into account the state-of-the-art archaeobotanical finds.  Table S2).
Based on the results of the SSAP analysis, a subset of 103 genotypes, including 37 T. araraticum collected from different geographic regions and representing all genetic groups, one T. militinae, 14 T. timopheevii, 38 T. dicoccoides, 9 T. dicoccon and four T. durum, was selected for a complementary analysis using Amplified Fragment Length Polymorphism (AFLP) markers to infer the population structure of GGA t A t wheats (Supplementary Table S2). Of them, forty-seven T. araraticum and T. timopheevii accessions (88.7%) were tested with SSAP and 39 (73.6%) with C-banding analyses.
Karyotype diversity of 370 T. araraticum accessions was assessed by C-banding and fluorescence in situ hybridization (FISH) in comparison with 17 T. timopheevii and one T. zhukovskyi genotypes (Supplementary Table S2). According to the C-banding analysis, most T. araraticum accessions (353 of 370) were karyotypically uniform and were treated as single genotype each. Seventeen accessions were heterogeneous and consisted of two (13 accessions) or even three (four accessions) cytogenetically distinct genotypes, which were treated as different entities (genotypes). In order to infer the population structure of GGA t A t wheats, 265 typical genotypes of T. araraticum were selected representing all karyotypic variants, seven T. timopheevii and one T. zhukovskyi. A total of 87 T. araraticum genotypes representing all geographic regions and chromosomal groups were selected for fluorescence in situ hybridization (FISH) analysis.

SSAP analysis
Evolutionary relationships among the comprehensive collection of wild and domesticated tetraploid wheat taxa were first inferred based on polymorphic retrotransposon insertions. For this, the highly multiplex genome fingerprinting method SSAP was implemented based on polymorphic insertions of retrotransposon families BARE-1 and Jeli spread across the wheat chromosomes. DNA was isolated from freezedried leaves of 787 SSD plants, using the Qiagen DNeasy Kit (Hilden, Germany). The SSAP protocol was based on Konovalov et al. (2010) with further optimizations for capillary-based fragment detection (Supplementary Material S3). In total, 656 polymorphic markers were generated for BBAA-and GGA t A t -genome wheats by amplification of multiple retrotransposon insertion sites. Data analysis was performed in SplitsTree 4.15.1 (Huson and Bryant 2006). NeighborNet planar graphs of Dice distances (Dice 1945) were constructed based on presence/absence of SSAP bands in the samples.

AFLP analysis
The AFLP protocol, as described by Zabeau and Vos (1993), was performed with minor modifications according to Altıntaş et al. (2008) and Alsaleh et al. (2015). In total, six AFLP primer combinations were used to screen the collection of 103 lines (Supplementary Material S3). Neighbor-joining (NJ) trees were computed based on Jaccard distances (Jaccard 1908;Perrier et al. 2003). NeighborNet planar graphs were generated based on Hamming distances (Huson and Bryant 2006). Genetic diversity parameters and genetic distances were calculated using Genalex 6.5 (Peakall and Smouse 2012).

C-banding
Chromosomal preparation and C-banding procedure followed the protocol published by Badaeva et al. (1994). The A t -and G-genome chromosomes were classified according to the nomenclature proposed by Badaeva et al. (1991) except for chromosomes 3A t and 4A t . Based on meiotic analysis of the F 1 T. timopheevii × T. turgidum hybrids (Rodríguez et al. 2000b) and considering karyotype structure of the synthetic wheat T. × soveticum (Zhebrak) (Mitrofanova et al. 2016), the chromosomes 3A t and 4A t were exchanged. Population structure of T. araraticum and the phylogenetic relationship with T. timopheevii were inferred based on chromosomal passports compiled for 265 genotypes representing all geographic regions and chromosomal groups (247 T. araraticum, 17 T. timopheevii and one T. zhukovskyi). Chromosomal passports were constructed by comparing the karyotype of the particular accession with the generalized idiogram of A t -and G-genome chromosomes ( Fig. 1; Supplementary Material S3), as described for T. dicoccon Schrank (Badaeva et al. 2015b).

Fluorescence in situ hybridization (FISH)
Two polymorphic G-genome-specific DNA probes, Spelt-1 (Salina et al. 1998) and Spelt-52 (Salina et al. 2004), were used to screen all 95 (87 T. araraticum, seven T. timopheevii and one T. zhukovskyi) genotypes considered for FISH analysis. Additionally, 26 of the 95 genotypes were analyzed using the probe pAesp_SAT86 (Badaeva et al. 2015a), which also showed differences between the accessions. The probes pSc119.2 (Bedbrook et al. 1980), GAA n (Pedersen et al. 1996) and pTa-535 (Komuro et al. 2013) were subsequently hybridized to the same chromosomal spread to allow chromosome identification. Classification of pSc119.2-and pTa-535-labeled chromosomes followed the nomenclature of Jiang and Gill 1993). Chromosomes hybridized with the GAA 10 microsatellite sequence were classified according to nomenclature suggested for C-banded chromosomes.

Genetic diversity and population structure of the GGA t A t genepool
Simultaneous amplification of multiple retrotransposon insertion sites using eight Long Terminal Repeat (LTR) primer combinations generated 656 polymorphic Sequence-Specific Amplification Polymorphism (SSAP) markers: 255 markers were obtained for Jeli insertions and 401 markers for BARE-1 insertions. According to our previous study (Konovalov et al. 2010), Jeli targets mostly the A genome, while BARE-1 is distributed between A-and B/G-genome chromosomes. Altogether, 787 wheat genotypes representing 753 genebank accessions were considered for data analysis, after excluding apparently misidentified taxa and several cases of low-yield DNA extraction (Supplementary  Table S2, column 6). Several major observations were made: (1) the extant GGA t A t genepool consists of three distinct lineages (two T. araraticum lineages and one of T. timopheevii, TIM). Surprisingly, wild T. araraticum consists of two major genetic lineages, preliminarily designated as 'ARA-0' and 'ARA-1' ( Table S6); (5) differences between A t − and A t + G genome diversity patterns based on Jeli and BARE-1 were discovered, respectively (Supplementary Table S6); and (6) Nei's genetic distance between lineages based on all 656 SSAP markers or considering only the BARE-1 markers revealed that ARA-0 is phylogenetically more closely related to TIM than ARA-1. However, considering only the Jeli markers, ARA-1 was more closely related to TIM. Triticum dicoccoides (DIC) was genetically related most closely to the ARA-1 lineage (Supplementary Table S6).
A carefully selected subset of 103 genotypes was used for Amplified Fragment Length Polymorphism (AFLP) analysis (Supplementary Table S2, column 8). A screen using the six most promising AFLP primer combinations uncovered a total of 146 polymorphic markers across all genotypes. Major findings can be summarized as: (1) clustering of genotypes based on AFLP markers corroborates the existence of the three major and genetically distinct lineages of GGA t A t wheats (Supplementary Figure S7). One lineage comprised all T. timopheevii (and its derived mutant T. militinae Zhuk. et Migusch.) genotypes. Importantly, the two lineages of T. araraticum and their distribution ranges were verified; (2) the summary statistics highlight that, in contrast to the SSAP analysis, ARA-1 was more diverse than TIM for all parameters (Supplementary Table S8); (3) ARA-1 was found to be genetically most closely related to TIM (Supplementary  Table S8); (4) ARA-1 was related most closely to DIC (Supplementary Table S8); (5) potential hybridization signals between ARA-0 and ARA-1 lineages were identified (Supplementary Figure S7, purple split); (6) ARA-1 lines collected around Kilis, Kahramanmaraş and Gaziantep in Turkey were genetically closest to TIM (Supplementary Figure  S7, purple split); and (7) shared splits between ARA-0 genotypes collected in Armenia and Azerbaijan with TIM (Supplementary Figure S7, purple split) indicate the potential Fig. 1 Generalized idiogram and nomenclature of the A t -and G-genome chromosomes. The C-banding pattern is shown on the left, the pSc119.2 (red) and pAesp_SAT86 (green) pattern on the right side of each chromosome. 1-7-homoeologous groups; S-short arm, L-long arm. The numerals on the left-hand side designate putative positions of C-bands/FISH sites that can be detected on the chromosome arm; C-bands specific for the ARA-1 group are shown with pink numerals, C-bands specific for the ARA-0 group are indicated by green numerals. Red asterisks on the right-hand side indicate C-bands that were considered for the 'chromosomal passport' contribution of ARA-0 to the formation of TIM or probably hybridization of ARA-0 with TIM. The genetically closest ARA-0 genotype to TIM was collected from the present Ararat province of Armenia.
In total, 248 genotypes of T. araraticum and 17 of T. timopheevii collected across the entire distribution range and representing all karyotypic variants (Supplementary  Table S2, column 12) were selected to infer the population structure based on chromosomal passports (Badaeva et al. 2015b). The results based on 96 informative C-bands supported the molecular findings using Jeli markers and AFLP markers and were congruent with the AFLP marker results as the closest ARA-1 genotypes to TIM were collected near Gaziantep (61 km SE from Türkoğlu, SW of Karadağ), Turkey  Table S10).
Fluorescence in situ hybridization (FISH) using six DNA probes was carried out for 95 genotypes (Supplementary  Table S2, column 13), which, in turn, corroborates the findings obtained using Jeli markers, AFLP and C-banding markers (Supplementary Table S11).

Intraspecific genetic diversity of T. araraticum based on karyotype structure and C-banding patterns
Cytogenetic analysis of 391 T. araraticum genotypes in comparison with 17 T. timopheevii genotypes provided detailed insights into the genetic composition of GGA t A t wheats. First, cytogenetic analysis highlighted significant differences between wild emmer T. dicoccoides and T. araraticum/T. timopheevii in karyotype structure and C-banding patterns (Fig. 4). Second, we revealed high diversity of the C-banding patterns and broad translocation polymorphisms within T. araraticum (Figs. 4,5). The karyotype lacking chromosomal rearrangements was defined as 'normal' (N) and it was the most frequent karyotype variant shared by T. timopheevii and T. araraticum. The 'normal' karyotype was found in 175 of 391 T. araraticum genotypes (44.6%) (more specifically: 155 of 342 ARA-0 = 45.32%; 20 of 49 ARA-1 = 40.81%). These 175 genotypes differed from each other only in the presence/absence or size of one to several C-bands. The ratio of karyotypically normal genotypes decreased from 86.7% in Azerbaijan (excluding Nakhichevan), to 60.0% in Turkey, 46.5% in Iraq, 30.6% in Armenia, 20.0% in Nakhichevan, Azerbaijan, 16.7% in Syria, to 9.1% in Iran (Supplementary Figure S12). Local populations differed in the ratio of normal/rearranged genotypes. For example, some populations from Dahuk, Iraq, possessed only karyotypically normal genotypes, while in others all genotypes possessed chromosomal rearrangements. Similarly, in Turkey the frequency of karyotypically normal genotypes varied from 100% (Mardin) to 0% (Kilis) (Supplementary Table S13).
C-banding patterns of T. araraticum were highly polymorphic. Based on the presence or absence of particular  Figure S20). Interestingly, genebank accession TA1900 presumably collected 32 km S of Denizli, Dulkadiroğlu district, Kahramanmaraş province in the Taurus Mountain Range of Turkey in 1959 shared karyotypic features of T. timopheevii and ARA-0 ( Fig. 4) and, probably, it is a natural or artificially produced hybrid of T. timopheevii with an unknown genotype of T. araraticum.
Only few C-bands were characteristic for the ARA-0 lineage (Figs. 1, 4). Three C-bands appeared with a frequency of over 95%, and two of them, 6A t L3 and 5GL15, occurred only in the ARA-0 group. The third C-band, 2GL13, was detected in 97% of ARA-0, but also in few ARA-1 genotypes.
'Region-specific' C-bands (in terms of highest frequency) (Badaeva et al. 1994) were detected for ARA-0. For example, (i) the band 4A t L7 dominated in Turkey and Transcaucasia (Supplementary Figure S14, Supplementary Figure S19); (ii) a medium to large 3A t S7 band was frequently observed in Iran and Sulaymaniyah (Iraq); (ii) one distinct 5GL13 band was frequently detected in genotypes from Erbil, Iraq (Supplementary Figure S15, Supplementary Figure S16). The specificity of C-banding patterns in genotypes originating from the same geographic region was not only determined by single bands, but usually by a particular combination of C-bands on several chromosomes. These region-specific banding patterns were observed for both, normal and translocated forms. For example, the unique banding pattern of chromosome 7G, lacking the marker C-band 7GS11, but carrying large bands for 7GS13, 7GS17 and 7GS21 on the short arm, and 7GL13 and 7GL15 on the long arm was common in Transcaucasia (Supplementary Figure S14). The chromosome 6G lacking telomeric C-bands on the long arm was frequent in genotypes from Dahuk and Sulaymaniyah (both Iraq), and also occurred in ARA-1 genotypes from Kahramanmaraş, Turkey (Supplementary Figure S20).
Altogether, we revealed 52 (33 novel) variants of single chromosomal rearrangements (Figs. 5, 6). They included paracentric (one variant) and pericentric inversions (seven variants) and 44 single translocations involving A t -A t , A t -G, or G-G-genome chromosomes. Double rearrangements were represented by 16 independent and three cyclic translocations; among them 10 were novel. Triple translocations were represented by three variants, two of which-T2A t :7G + T6A t :5G:6G and T3G:7G:7A t + T6A t :6G were found here for the first time. Both variants of quadruple translocations have been identified earlier in Transcaucasia (Badaeva et al. 1990(Badaeva et al. , 1994. A translocation between two chromosomes could give rise to different products depending on the breakpoint position and arm combination in rearranged chromosomes. For example, a centromeric translocation between 1G and 2G resulted in two translocation variants which were distinct in arm combinations (S:S vs. S:L). Three translocation variants involving chromosomes 3G and 4G differed from each other in arm combination and breakpoint position. To discriminate  KU-1901 (Armenia). 1-7-homoeologous groups. C-bands typical for ARA-1 are indicated with blue arrows, for ARA-0-with green arrows, and C-bands characteristic for T. timopheevii-with red arrows. Black arrows point to rearranged chromosomes in genotypes different translocation variants involving same chromosomes, they were designated as T1G:2G-1 and T1G:2G-2, etc. (Supplementary Table S13).
Most variants of chromosomal rearrangements were unique and identified in one or few genotypes, and only four variants were relatively frequent. These were a triple  Supplementary Table S13. Novel variants are designated with black numbers, and already known variants by red numbers translocation T2A t :4G:7G + T4A t :7A t , perInv7A t -1, T6G:7G and T2G:4G:6G (21, 16, 12 and 13 genotypes, respectively). Taken together, these four variants accounted for approximately 16% of the whole materials we studied. Genotypes carrying the same rearrangement usually had similar C-banding patterns and were collected from the same, or closely located geographic regions. For example, (1) all genotypes with T2A t :4G:7G + T4A t :7A t originated from Nakhichevan, Azerbaijan; (2) T6G:7G and T2G:4G:6G were found in Erbil, Iraq; and (3) perInv7A t -1 was collected in Iran and in the neighboring region of Sulaymaniyah, Iraq.
Other frequent translocations had a more restricted distribution and usually occurred in a single population. Only few genotypes carrying the same translocations were identified in spatially separated populations. These genotypes differed in their C-banding patterns, for example: (1) T2G:4G was found in four genotypes from Erbil, Iraq and in two genotypes from Siirt, Turkey; (2) T1G:3G was identified in Combinations of chromosome arms in rearranged chromosomes are designated. Line colors mark the different groups: ARA-0 (green), ARA-1 (pink) and T. timopheevii (black). Solid arrows designate novel rearrangements; arrows with asterisk designate previously described rearrangements (Badaeva et al. 1990(Badaeva et al. , 1994. The numerals above/next to the arrows indicate the number of accessions carrying the respective translocation five ARA-0 genotypes from Dahuk, Iraq, one ARA-1 from Turkey and one ARA-1 genotype of a mixed accession IG 117895 collected in Syria; and (3) T1G:5G was identified not only in three cytogenetically distinct T. araraticum ARA-0 genotypes from Armenia and Azerbaijan and ARA-1 from Turkey ( Fig. 6; Supplementary Figure S14, Supplementary Figure S20), but also in T. timopheevii .
Most populations consisted of both genotypes with 'normal' karyotype and genotypes with one to several variants of chromosomal rearrangements. However, their ratio and spectra differed between regions ( Fig. 6; Supplementary Figure S12; Supplementary Material S21).

Cytogenetic diversity of GGA t A t wheats assessed using FISH markers
To further investigate the intraspecific diversity of T. araraticum and to assess their phylogenetic relationships with T. timopheevii, we carried out FISH using six DNA probes. The probes pTa-535, pSc119.2 and GAA n ensured chromosome identification , whereas Spelt-1, Spelt-52 and pAesp_SAT86 were used to estimate intra-and interspecific variation (Fig. 7).
The variability of pAesp_SAT86 hybridization patterns was analyzed in four T. timopheevii and 26 T. araraticum genotypes, of them seven were from ARA-1 and 19 from ARA-0 lineages (Supplementary Table S2). Distribution of the pAesp_SAT86 probe in all T. timopheevii genotypes was similar except for chromosome 7A t in genotype K-38555, which was modified due to a paracentric inversion or insertion of an unknown chromosomal fragment. Labeling patterns, however, were highly polymorphic for T. araraticum ( Fig. 7; Supplementary Figure S22). Large pAesp_SAT86 sites were found only on some G-genome chromosomes. The A t -genome chromosomes possessed several small, but genetically informative polymorphic sites. Some of these sites were lineage-specific. Most obvious differences were observed for 3A t , 4G and 7G chromosomes (Fig. 1). Thus, all TIM and ARA-1 genotypes carried the pAesp_SAT86 signal in the middle of 3A t S, while for ARA-0, it was located sub-terminally on the long arm. One large pAesp_SAT86 cluster was present on the long arm of 4G in ARA-0, but on the short arm in ARA-1 and TIM. Two large and adjacent pAesp_SAT86 clusters were detected on 7GS in ARA-1 and TIM, but they were split between opposite chromosome arms in all ARA-0 genotypes. Differences between ARA-0 and ARA-1 in pAesp_SAT86 cluster position on 4G and 7G could be caused by pericentric inversions. Some other pAesp_SAT86 sites were identified in either one of the three groups. ARA-1 exhibited the largest polymorphism of pAesp_SAT86 labeling patterns among all groups (Supplementary Figure S22).
FISH with Spelt-1 and Spelt-52 probes on chromosomes of 87 T. araraticum genotypes from different chromosomal groups and of different geographic origin, seven T. timopheevii, and one T. zhukovskyi genotypes revealed high intraspecific diversity of T. araraticum and low polymorphism in T. timopheevii (Supplementary Figure S23). The broadest spectra of labeling patterns were found in genotypes from Dahuk and Sulaymaniyah (Iraq) and in the ARA-1 group from Turkey, while material from Transcaucasia exhibited the lowest variation. The polymorphism was due to variation in the number of Spelt-1 and Spelt-52 sites, their size and chromosomal location (Supplementary Table S24). The Spelt-1 signals in various combinations appeared in sub-telomeric regions of either one or both arms of 2A t , 6A t , and all G-genome chromosomes (Zoshchuk et al. 2007). The Spelt-52 signals were observed in various combinations on 2A t S, 1GS, 2GS and 6GL chromosomes. Among them, two Spelt-1 sites (6GL and 7GL) and one Spelt-52 locus (2GS) were novel. The patterns of Spelt-1 and Spelt-52 repeats varied across geographic regions and between chromosomal groups (Supplementary Material S25; Supplementary Table S26).

Discussion
We present the most comprehensive survey of cytogenetic and genomic diversity of GGA t A t wheats. The composition, distribution and characteristics of the GGA t A t genepool are described. Building on our results, the latest published complementary genomics studies and state-of-the-art archaeobotanical evidence we revisit the domestication history of the GGA t A t wheats. We arrived at the following four key findings:

The GGA t A t genepool consists of three distinct lineages
We sampled the full breadth and depth of GGA t A t wheat diversity and discovered a clear genetic and geographic differentiation among extant GGA t A t 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 southeastern Turkey and in neighboring northwestern 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. Fig. 3, collecting gaps are evident for T. araraticum, and future collecting missions should focus on four specific regions: (1) 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; (2) between Bitlis (Turkey)-Amadiyah (Iraq) in the west and the Armenian border in the east including northwestern Iran; (3) between Adiyaman-Silvan in the We found only two T. araraticum populations which contained representatives of both ARA lineages: (1) 45 km southeast of Kahramanmaraş to Gaziantep, and (2) 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.

Based on
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 southeastern 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 GGA t A t 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 geographic 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 GGA t A t genepool taxonomically in the future.
2. The karyotypic composition of GGA t A t wheats is as complex as the phylogenetic history of the GGA t A t genepool Based on C-banding polymorphisms, 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 '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 1A t L3, 4A t S7, 5A t L3, 1GL5, 2GL7, 3GL7 + L11 (Fig. 1). As expected, the number of C-bands characteristic for ARA-0 was smaller (due to the wide geographic distribution) and only two C-bands were lineage-specific and found in normal as well as translocated genotypes: 6A t L3 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: (1) TIM and ARA-1 carry the pSc119.2 signal in the middle of 1A t long arm, while this site was absent from ARA-0; (2) 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; (3) the pAesp_SAT86 patterns on chromosomes 3A t , 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 A t (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. (2017Bernhardt et al. ( , 2020. Regular chromosome pairing observed in the F 1 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 7A t 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 registered as T. araraticum. Relatively good chromosome pairing was observed in the F 1 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.

Does the T. timopheevii population found in western
Georgia represent the last remnant of a widespread ancient cultivation area of GGA t A t wheats?
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(Stoletova -1925Zhukovsky 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 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).
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(Bogaard et al. , 2017Ergun 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. 7100-5500 cal BC) (Bogaard et al. 2013(Bogaard et al. , 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 archeological finds of NGW . 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 GGA t A t genepool . 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 GGA t A t wheat in general) was domesticated from T. araraticum during early agriculture, and was widely cultivated in the prehistoric past .
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 GGA t A t 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 northeastern Turkey. Menabde and Ericzjan (1942;cited by Dorofeev et al. 1979) 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 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, three T. timopheevii accessions maintained in two ex situ genebanks are reported to potentially originate from Turkey (Supplementary Table S2 (Schiemann 1934); and (2) 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; and (3) KU-107-5 maintained in a genebank of Kyoto University, Japan, was obtained from a collection of Ankara University, Turkey, in 1960. The exact collection site of this accession, however, is not known. Two accessions, PI 119442 and ATRI 3433 harbor the 'normal' karyotype of T. timopheevii, and all three 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 and Genomics Resource Center, Kansas State University, USA, 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 are correct, we could speculate that T. timopheevii may have been cultivated in Turkey during the first half of the twentieth century. However, is this realistic option?
We believe not. As reported by Stoletova (1924Stoletova ( -1925, 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 and Hijmans 2017;R Core Team 2016) 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(Davis -1988Hanelt 2001). From this, we conclude that the cultivation range of T. timopheevii (s str.) was not wider in the recent past.

Conclusions
The evolutionary history of the GGA t A t wheat genepool is complex, as several chromosomal rearrangements were involved. Since different types of markers can lead to different results, it is crucial to consider all available data sets and to use several types of markers to sketch a holistic picture.
However, some pieces of the puzzle are clearly recognizable: the region around Dahuk in northern Iraq can be considered the center of origin, but also the center of diversity of T. araraticum.
The origin of T. timopheevii s.str. remains unclear, but we speculate that it was probably introduced from Turkey, on the grounds that wild T. araraticum does not grow in Georgia and that T. timopheevii s.str. is more closely related to T. araraticum from Turkey than to the Transcaucasian types. Based on bioclimatic variables, we predict that the remnant T. timopheevii s.str. is maladapted to the climate outside Western Georgia. If this speculation is correct, it suggests a sister-group relationship between (1) the Georgian T. timopheevii s.str. and both T. araraticum lineages (ARA-0, ARA-1), but also between (2) T. timopheevii s.str. and the prehistoric SW Asian T. timopheevii s.l. (= NGW). In other words, it is not clear yet, whether T. timopheevii s.str. found in western Georgia was a direct descendant of T. timopheevii s.l., or whether the T. timopheevii s.str. was a product of local domestication (in Northern Turkey) independent of the domestication of T. timopheevii s.l. To answer this question, comparative analyses at the whole genome level need to be conducted (Keilwagen et al. 2019;Wendler et al. 2015;Walkowiak et al. 2020).
The oldest known records of prehistoric T. timopheevii s.l. are of Turkish origin: (1) Aṣıklı Höyük in Cappadocia, and (2) Cafer Höyük (just inside the Fertile Crescent and potentially older than Aṣıklı Höyük but less thoroughly dated) (Cauvin et al. 2011;Quade et al. 2018), though more prehistoric finds may be identified in the future. Emphasis for further archaeobotanical research should be given to the central part of the Fertile Crescent, including northern Iraq.
The unexpected discovery of the ARA-1 lineage is exciting and may have implications for our understanding of the origins of agriculture in southwest Asia. Was ARA-1 involved in the formation of T. timopheevii s.l. (= NGW)?
The distribution area of the ARA-1 lineage requires our attention. It is interesting to note that the ARA-1 lineage (GGA t A t ) and T. dicoccoides (BBAA) grow in sympatry and in mixed populations only in a peculiar geographic area in the Northern Levant. This specific area is located between Kahramanmaraş in the north, Gaziantep in the east, Aleppo in the south and the eastern foothills of the Nur Dağlari mountains in the west. It is interesting that the closest extant wild relatives of domesticated einkorn, barley and emmer were also collected here: (1) the beta race of wild einkorn (Kilian et al. 2007) and the closest wild relatives to einkorn btr1 type (Pourkheirandish et al. 2018); (2) the closest wild barley to btr2 barley (Pourkheirandish et al. 2015) and to btr1b barley (Civáň and Brown 2017). This specific region was among the areas predicted with high probability as potential refugia for wild barley during the Last Glacial Maximum (LGM) by Jakob et al. (2014); and (3) wild emmer subgroup II of the central-eastern race (Özkan et al. 2011). On the other hand, some genetic research has suggested that domesticated emmer wheat and barley received substantial genetic input from other regions of the Fertile Crescent, resulting in hybridized populations of different wild lineages indicating a mosaic of ancestry from multiple source populations (Civáň et al. 2013;Oliveira et al. 2020;Poets et al. 2015). It will be interesting to see whether further genetic and archaeobotanical research on T. araraticum lineages and T. timopheevii s.l. can help to resolve this issue.
Finally, and fortunately, 1294 accessions of T. araraticum and 590 accessions of T. timopheevii s.str. are stored in 24 and 37 ex situ genebank repositories, respectively (Knüpffer 2009). Our study provides the basis for a more efficient use of T. araraticum and T. timopheevii materials for crop improvement.

Conflicts of interest
The authors declare that they have no conflict of interest.

Ethical approval
The authors declare that the experiments comply with the current laws of Germany.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.