Evolutionary History of Triticum petropavlovskyi Udacz. et Migusch. Inferred from the Sequences of the 3-Phosphoglycerate Kinase Gene

Single- and low-copy genes are less likely to be subject to concerted evolution. Thus, they are appropriate tools to study the origin and evolution of polyploidy plant taxa. The plastid 3-phosphoglycerate kinase gene (Pgk-1) sequences from 44 accessions of Triticum and Aegilops, representing diploid, tetraploid, and hexaploid wheats, were used to estimate the origin of Triticum petropavlovskyi. Our phylogenetic analysis was carried out on exon+intron, exon and intron sequences, using maximum likelihood, Bayesian inference and haplotype networking. We found the D genome sequences of Pgk-1 genes from T. petropavlovskyi are similar to the D genome orthologs in T. aestivum, while their relationship with Ae. tauschii is more distant. The A genome sequences of T. petropavlovskyi group with those of T. polonicum, but its Pgk-1 B genome sequences to some extent diverge from those of other species of Triticum. Our data do not support for the origin of T. petropavlovskyi either as an independent allopolyploidization event between Ae. tauschii and T. polonicum, or as a monomendelian mutation in T. aestivum. We suggest that T. petropavlovskyi originated via spontaneous introgression from T. polonicum into T. aestivum. The dating of this introgression indicates an age of 0.78 million years; a further mutation event concerning the B genome occurred 0.69 million years ago.

The origin of T. petropavlovskyi has been discussed for decades. Gorsky [6] analyzed the morphology of Xinjiang rice wheat, and suggested that it was a mutated form of the tetraploid Triticum polonicum L.. However, Udachin and Miguschova [7] discovered that the Xinjiang rice wheat is not tetraploid but hexaploid, and named it T. petropavlovskyi Udacz. et Migusch. The chromosomal constitutions of the Xinjiang rice wheat is AABBDD [2,[8][9][10]. Dorofeev et al. [11] hypothesized that T. petropavlovskyi could be the result of spontaneous hybridization between T. aestivum and T.
polonicum. Several previous studies indicated that the genes supporting a long glume in T. polonicum and T. petropavlovskyi were allelic and located on the long arm of chromosome 7A [12][13][14], agreeing with the hypothesis of spontaneous hybridization. Yen et al. [15] studied the natural distribution in Xinjiang of Aegilops tauschii, and Yang et al. [16] and Liu et al. [17] reported a similar distribution for a dwarfing accession of T. polonicum and suggested that T. petropavlovskyi is derived from a hybridization event between T. polonicum and Ae. tauschii. However, Efremova et al. [18] maintained that T. petropavlovskyi originated from T. aestivum through spontaneous mutation.
Despite prior intensive research, the origin of T. petropavlovskyi is still uncertain, and three hypotheses have been proposed: (1) T. petropavlovskyi is an independent species is derived from a natural hybridization event between T. polonicum and Ae. tauschii [10,15,[19][20][21]; (2) T. petropavlovskyi is a natural cross or backcross between T. polonicum and T. aestivum [2,11,12,22,23]; and (3) T. petropavlovskyi is a monogenic mutant of T. aestivum [18,24]. In a recently study, Kang et al. [25] created the synthetic hexaploid wheat (SHW-DPW) between T. polonicum from Xinjiang and Ae. tauschii: its spike  morphology was similar to T. petropavlovskyi. However, a comparison of SHW-DPW to T. petropavlovskyi, T. polonicum and related species by the phylogenetic analysis of the Acc-1 gene indicated that T. petropavlovskyi originated from the cross between T. polonicum from Xinjiang and exotic landraces of T. aestivum [26]. This finding contradicts the morphology-based conclusion of preceding study. Previous works based on different methods, including cytology, morphology and nuclear markers, have failed in identifying the origin of T. petropavlovskyi. Furthermore, no molecular-clock has been reported to examine the timing of its origin. Single-and low-copy nuclear genes, being less susceptible to concerted evolution [27][28][29], are useful in phylogenetic study [30][31][32][33] as well as in the identification of parents of allopolyploidy taxa [26,[34][35][36]. Genes such as acetyl-CoA carboxylase 1 (Acc-1) [30], disrupted meiotic cDNA 1 (DMC1) and translation elongation factor G (EF-G) [37] have been particularly useful in elucidating the phylogenesis of Triticum-Aegilops species. The plastid 3phosphoglycerate kinase (PGK) gene, Pgk-1, is a single copy nuclear gene in diploid species of the Triticeae; it is frequently considered to be superior to the Acc-1 gene for assessing the evolutionary history of polyploid wheats, because the Pgk-1 gene has more parsimony informative sites than the Acc-1 gene [31,38]. For allotetraploid and allohexaploid species with two or three copies of genes present as single copies in diploid ancestors, the Pgk-1 gene can both elucidate the phylogenetic relationships of such polyploid as well as potential progenitors [30,31,39,40].
In this study, we sequenced and analyzed the single-copy nuclear Pgk-1 gene in the following taxa: T. petropavlovskyi, SHW-DPW (synthetic hexaploid wheat between T. polonicum and Ae. tauschii), and the hypothetical Triticum and Aegilops progenitors of T. petropavlovskyi to reveal their phylogenetic relationships and to explore both the origin of T. petropavlovskyi and its divergence time from other taxa.

Plant materials
The species, genomic constitutions, origin, GenBank accessions, and sources of the taxa are listed in Table 1. The sequences of pgk-1 gene of the accensions with TA numbers were obtained from the GenBank database; the rest of the species considered for the sequences are reported here for the first time.
The accessions with PI and AS numbers were kindly provided by the American National Plant Germplasm System (Pullman, Washington, USA) and the Triticeae Research Institute, Sichuan Agricultural University, China, respectively. The artificial synthetic amphiploid of Triticum polonicum and Aegilops tauschii (SHW-DPW) was produced by Kang et al. [25]. The plants and voucher specimens have been deposited at Herbarium of Triticeae Research Institute, Sichuan Agricultural University, China (SAUTI).

DNA extraction, amplification and sequencing
DNA was extracted from fresh leaves of single plants, following a standard CTAB (cetyltrimethylammonium bromide) protocol [41]. For amplification of the Pgk-1 gene, a pair of Pgk1-specific primers, PPF1 (59-CACCTGGGTCGTCCTAAGGGTGTT-39) and PPR1 (59-ACCACCAGTTGTGTTGTGGCTCAT-39), was used [31]. Polymerase chain reactions (PCR) were performed in a GeneAmp 9700 Thermal Cycler (Applied Biosystems Inc., California, USA) according to the following cycling program: initial denaturation at 94uC for 5 min; 35 cycles of 94uC for 30 s, 56uC for 30 s, 68uC for 5 min; followed by a final elongation period at 68uC for 10 min. A final volume of 50 ml for each PCR reaction was prepared, containing 0.5 mg of genomic DNA, 106 reaction buffer, 1.5 mM of each primer, 2.5 mM of each dNTP, 2.5 mM MgCl2, 2 units of high-fidelity ExTaq DNA polymerase (Takara Biotechnology Co. Ltd., Dalian, China).
1.0% agarose gel was used to estimate the size of the amplification products, which were purified using the EZNATN gel extraction kit (Omega Bio-Tech, Georgia, USA) and stored in 30 ml TE buffer. The purified products were cloned into the pMD19-T vector (Takara) according to the manufacturer's instructions. Cloning of PCR amplifications from single-copy nuclear genes from allopolyploid species should isolate homoeologous sequences from each nuclear genome [42,43]. For the hexaploid Triticum species, the A, B and D genomes homoeologous sequences of Pgk-1 gene were isolated, and the A and B genomes homoeologous sequences were separated for the tetraploid Triticum. The cloned PCR products were commercially sequenced on both strands by the Beijing Genomics Institute (BGI, Shenzhen, China). All the sequences used in the phylogenetic analysis were derived from at least five independent clones.

Alignments and phylogenetic analysis
Multiple sequences were aligned using Clustal X with default parameters, followed by manual adjustment to minimize gaps [44]. In an initial phylogenetic analysis, if all sequences used for alignment derived from independent clones, formed a monophyletic group, then only one sequence was used later on. Distinct sequences derived from single accessions mapping different clades were all included in the phylogenetic analysis. Nucleotide frequencies, transition/transversion ration, and variability in different regions of the sequences were examined by MEGA 5.0 [45].
Three data matrixes, including exon+intron data (the target Pgk-1 gene sequences), exon data and intron data, were used separately to implement phylogenetic analyses. Phylogenetic trees were Evolutionary History of Triticum petropavlovskyi created using Maximum likelihood (ML) and Bayesian inference (BI). ML analysis was carried out with PAUP*4.0b10 (Swofford, D. L., Sinauer Associates, http://www.sinauer.com), and Psathyrostachys juncea (Fischer) Nevski was used as an outgroup. The best-fit models of sequence evolution for ML analysis were estimated using ModelTest v3.0 with Akaike information criteria (AIC) [46]. The optimal models were found to be HKY+G for the exon+intron data, TVM+G for the intron data, and TrN+G for exon data. ML heuristic searches were performed with 100 random addition sequence replications and TBR branch swapping algorithm. The robustness of the trees was estimated using bootstrap support (BS) [47]. ML bootstrapping was performed with 250 replicates, each with three replicates of stepwise random taxon addition, using the same model and parameters. BS-values under 50% were not included in figures. Bayesian inference (BI) analysis was performed using MrBayes v3.2 [48] with Psathyrostachys juncea used as an outgroup. The bestfit models for BI analysis were carried out with AIC using MrModelTest v2.3 (http://www.ebc.uu.se/systzoo/staff/ nylander.html). The optimal models were found to be GTR+I+G for exon+intron data, GTR+G for intron data and exon data. Four MCMC (Markov Chain Monte Carlo) chains (one cold and three heated) were applied with default setting. In order to make the standard deviation of split frequencies fall below 0.01, 4,200,000 generations for exon+intron data, 3,000,000 generations for intron data, and 5,000,000 for exon data were run. Samples were taken every 100 generations under the best-fit model. For all analyses, the first 25% of samples from each run were discarded as ''burn-in'' to ensure the stationarity of the chains. Bayesian posterior probability (PP) values were obtained from a majority rule consensus tree generated from the remaining sampled trees. PP-values less than 90% were not included in figures.

Network analysis
The median-joining (MJ) network method [49], and Templeton, Crandall and Sing (TCS) method [36], have been shown to be effective methods for revealing specific progenitor descendant relationships of perennial Triticeae [21,[50][51][52], and were thus performed in this study. Before reconstructing the MJ and TCS networks, a test of recombination was performed using the Phi (Pairwiser homoplasy index) method within Splits tree [53]. Building upon this test, the sequences of A (P = 0.64932), B (P = 0.9999) and D genomes (P = 1.0) were used to generate the MJ network, and the exon data (P = 0.7493) was used to generate the TCS network. MJ network analysis was generated by Network 4.1.1.2 program (Fluxus Technology Ltd, Clare, Suffolk, UK), and TCS haplotype network was performed to evaluate possible genetic relationships between haplotypes with the computer program TCS 1.2.1 [54].

Divergence dating
The potential clock-like evolution of Pgk-1 sequences was evaluated with a likelihood ratio rate comparing the likelihood scores from the unconstrained and clock-constrained analyses, implemented in PAUP*4.0b10. The molecular clock was rejected because the substitution rates were significantly heterogeneous (x2 = 154.90, df = 95, P = 0.0001), implying a very poor fit to the molecular clock. Therefore, divergence times with 95% confidence intervals (C.I.) were estimated using the Bayesian relaxed molecular clock method, implemented in BEAST v1.7.1 [55].
The lack of fossils for Triticeae precluded a direct calibration of tree topologies. Instead, molecular dating of the intron data was estimated on the basis of the intron region of the Pgk-1 gene clock of 0.0051 substitutions per site per MY (million year) [31,38], setting the clock for the divergence of Pooideae and Panicoideae sub-families at 60 MYA [56][57][58]. Calibration points were performed using a relaxed uncorrelated lognormal molecular clock. MCMC searches were run for 10,000,000 generations under GTR+G model (with the associated parameters specified by MrModelTest as the priors), with the first 2,000,000 discarded as burn-in. Trees were then viewed in FigTree v1.3.1 (http://tree. bio.ed.ac.uk/).

Sequence analysis
In all polyploid species considered, the expected number of copies of the Pgk-1 gene were successfully amplified. The DNA sequence of the Pgk-1 gene includes 5 exons and 4 introns, which range in length from 1360 bp to 1476 bp, as known from previous study [30,31,38] (Table 2). The lengths of exon+intron, exon and intron data sets were 1466, 894 and 572 bp, respectively. As expected, the level of nucleotide variation in the exon region (88 variable sites and 36 parsimony-informative sites) was lower than that in the intron region (107 variable sites and 80 parsimonyinformative sites). The average content of G+C of exon+intron, and exon was 42.9, and 47.3%, respectively, and the transition/ transversion ratio was 2.13, and 2.24, respectively. The alignment of the exon sequence was unambiguous and without gaps. Gaps were, on the contrary, present in introns. In particular, apart from single nucleotide substitutions and deletions, three significant indels (insertion/deletion) (indel 1, 2, 3) were found (Fig. 1). Indel 1 was located at position 67-72 of the A genome; indel 2 mapped at position 563-568 and had a 6 bp deletion specific for A genome. Unexpectedly, the T. aestivum ssp. yunnanense (AS338) did not show the indel 2. Indel 3 was presented at position 1220-1308 and had a 89 bp insertion in the G genome of T. timopheevii.

Phylogenetic analyses
Using Psathyrostachys juncea as an outgroup, the three data sets corresponding to exon+intron, exon and intron were used phylogenetic analyses (ML and BI) were carried out. ML analysis of the exon+intron data generated a single phylogenetic tree (2Lnlikelihood = 4092.85), with the following parameters: A = 0.26; C = 0.21; G = 0.23, T = 0.30, gamma shape parameter = 0.29. Bayesian analysis of the same data recovered the same topology. In Figure 2, the ML tree is reported with values of the bootstrap support (BS) above and posterior probabilities (PP) below branches.
The ML tree of Figure 2 indicates that all homoeologous Pgk-1 sequences from polyploid accessions are grouped with those of the Evolutionary History of Triticum petropavlovskyi diploid parental species. The tree has two major clades: the one including sequences of the A genome and the second those derived from the genomes B, D, G and S. In Clade I, the A genome specific sequences from three T. petropavlovskyi accessions and from Triticum species (except T. aestivum cv. Chinese Spring) formed a group with 90% bootstrap value and 100% posterior probabilities support. One T. petropavlovskyi accession (AS358), together with two accessions of T. polonicum (AS304 and PI42209), formed a subclade, with bootstrap value of 95%. In Clade II, the B genome sequences from three T. petropavlovskyi accessions clustered together in a well supported (71% BS and 98% PP) subclade. SHW-DPW had a topology contiguous with two accessions of T. polonicum (AS302 and AS304), with 63% bootstrap support. The sequences from the D genomes mapped to two subclades. One consisted of three accessions of T. petropavlovskyi, eleven of T. aestivum and one of Ae. tauschii (TA1691), with 72% bootstrap support. The second one included SHW-DPW and Ae. Tauschii (AS60) with 87% BS and 100% PP.
ML analysis of the intron data yielded a single phylogenetic tree (2Lnlikelihood = 1804.87), with parameters: A = 0.29; C = 0.18; G = 0.17; T = 0.36 and gamma shape parameter = 0.60. The Bayesian analysis generated the same topology, as illustrated in Figure 3. Two major clades are evident: Clade I includes only the Pgk-1 sequences from the G genome with a high bootstrap support (99% BS, 98% PP). The Clade II includes A, B, D and S genomes sequences and is congruent with the tree inferred from the exon+intron data, except for nodes presenting different statistical support. In Clade II, sequences from the B genome clustered together with a good support (70% BS and 100% PP). In this B genome clade, with the exception of T. sphaerococcum (PI70711), all accessions were grouped in a well supported subclade (92% BS and 100% PP); the three accessions of T. petropavlovskyi were separated from other Triticum. In the A genome subclade, T. petropavlovskyi (AS359) mapped together with T. polonicum (AS304) (78% BS and 100% PP).
Exon data generated a single ML phylogenetic tree (2Lnlikelihood = 2150.29), with the following parameters A = 0.24, C = 0.21, G = 0.27, T = 0.28, gamma shape parameter = 0.38. ML and BI analysis of the same data supported a similar topology. Figure 4 reports the ML tree of exon sequences which includes two major clades: Clade I, consisting of A genome sequences: T. carthlicum (PI532509) and T. petropavlovskyi (AS360) are mapped in the same group (64% BS and 95% PP). In the second clade (Clade II), which includes B, D, G, and S genomes, the B genome sequences of three accessions of T. petropavlovskyi grouped together (64% BS and 93% PP), separated from other accessions. Ae. tauschii (TA1691) and Ae. speltoides (TA2368) clustered in a subclade (76% BS and 97% PP).

Network analysis
To highlight the relationships among haplotypes of the Pgk-1 sequence, network methods were employed and the exon+intron data (Fig. 5) and exon data (Fig. 6) were considered. In Figure 5, each circular network node represents a haplotype, with node size being proportional to number of its isolates. Mv (median vectors representing missing intermediates) shows unsampled nodes inferred from the MJ network analysis. The number on the branches indicates the positions of the mutations. Network loops represent either true reticulation events or alternative genealogies in closely related lineages. MJ analysis of the Pgk-1 exon+intron data recovered groupings corresponding to clades revealed by ML phylogeny. T. petropavlovskyi, as expected, was present in three clusters (A, B and D), representing the A, B and D genomes. Most accessions of T. aestivum, except T. aestivum cv. Chinese spring, were included in the A-type. The A-type sequences of three accessions of T. petropavlovskyi were included in subgroups I and II. T. petropavlovskyi (AS358) and T. polonicum (AS302) were placed at a central branching point. Meanwhile, in the B-type cluster, three accessions of T. petropavlovskyi grouped together in subgroup III, and T. polonicum (AS304) together with SHW-DPW in subgroup IV. In the D-type cluster, T. petropavlovskyi (AS358 and AS359), T. aestivum cv. Chinese Spring, T. aestivum cv. J-11, T. aestivum ssp. yunnanense (AS343) and T. spelta (PI347852) resulted included in subgroup V, while the sequences from the amphiploid SHW-DPW and Ae. tauschii formed the distinct subgroup VI. The TCS procedure [36] was used to illustrate haplotype relationships among accessions. TCS defined a 95% parsimony connection limit of 13 steps for exon alignment of fifty haplotypes derived from 96 sequences (Fig. 6). The TCS network consisted of three major haplotypic groups corresponding to the A, B and D genomes. The length of the branches between two nodes was proportional to the nucleotidic difference. In TCS analysis, T. petropavlovskyi (AS360) shows a close haplotype relationship with T. carthlicum from Xinjiang, China, supporting the exon results of the ML and BI analyses. Two further differences between the TCS and MJ were noted. Firstly, the haplotype of the D genome of T. macha (PI278660) appeared related to haplotypes of the B genome accessions. Secondly, Ae. tauschii (TA1691) showed a close haplotype relationship with the S genome of Ae. speltoides var. ligustica.

Molecular dating
The BEAST analysis of the intron region of Pgk-1 was used to derive a time-calibrated phylogenetic tree (Fig. 7). Under a lognormal relaxed clock, rate variation was equal to 0.96 (95% C.I., 0.67-1.39), supporting the adoption of the relaxed clock method. The Yule prior was equal to 0.47 (95% C.I., 0.37-0.63) and five homoeologous types of the Pgk-1 gene, A-, B-  Evolutionary History of Triticum petropavlovskyi the tree, the tetraploid wheat T. polonicum diverged earlier than T. petropavlovskyi. In the A genome, the divergence time of T. petropavlovskyi (AS358) was later than the other two accessions of T. petropavlovskyi. On the contrary, in B and D genomes, the divergence time was earlier than the other two accessions. Additionally, the divergence time of B genome of T. petropavlovskyi was earlier than other three Chinese endemic wheat landraces. Between T. petropavlovskyi and SHW-DPW, a significant difference was observed: in the genomes A, B and D, the divergence time of SHW-DPW was later than T. petropavlovskyi.

Relationships between T. petropavlovskyi and hexaploid wheat taxa
Based on cytological results, Yao et al. [3] and Chen et al. [23] suggested that the B genome was responsible for the difference between T. petropavlovskyi and T. aestivum cv. Chinese Spring, and that two pairs of chromosomes, one identified as chromosome 6B [2], were involved. The allelic variation at the HMW glutenin subunits loci, Gli-1 and Gli-2, supported the cytological results [59]. Also, Yang et al. [10] reported that T. petropavlovskyi differed from T. spelta in at least one or two pairs of chromosomes. Results based on molecular markers, including A-PAGE, SDS-PAGE, STS-PAGE, SSR and RFLP, indicate that T. petropavlovskyi is genetically distinct from other Chinese endemic wheat landraces [5,59].
Our ML and BI study of the Pgk-1 gene indicates that the A and D genomes of T. petropavlovskyi are basically shared with T. spelta, T. compactum and T. sphaerococcum. When the B genome is considered, T. petropavlovskyi groups in one subclade, comparatively distantly related to T. aestivum. In addition, the B-type of MJ network shows that the accessions of T. petropavlovskyi (subgroup III) are distinct from those of other species. SHW-DPW, a synthetic hexaploid wheat with both genomes of T. polonicum and Ae. tauschii [26], based on the Pgk-1 gene is characterized by the A, B and D genomes, the SHW-DPW is distant from those of T. petropavlovskyi in B and D genomes.

Relationships between T. petropavlovskyi and the tetraploid wheats
Morphologically, the spikelet of T. petropavlovskyi are similar to those of T. turanicum and T. polonicum [7,26]. Moreover, the cytology of interspecific hybrids between T. petropavlovskyi and tetraploid wheats support a closer relationship with the AABB genomes, compared to wheats with the AAGG genome [2]. According to Akond and Watanabe [24], T. petropavlovskyi is more closely related to T. polonicum than to T. durum or T. turgidum. However, Arbuzova et al. [60] and Efremova et al. [18] report that the genes supporting the elongated glumes in T. polonicum and T. petropavlovskyi are not allelic.
In the present study, based on the ML and BI analyses of the genome A Pgk-1 gene, two accessions of T. polonicum have a common topology with T. petropavlovskyi, while the TCS analysis of exon data indicates that the haplotype of A genome in T. petropavlovskyi (AS360) is more closely related to T. carthilicum than to other wheats. The phylogenetic analyses and MJ network specific for the B genome shows that three accessions of T. petropavlovskyi group together, and are topologically distant from those of tetraploid wheats. This finding indicates that the B genome of T. petropavlovskyi diverge from the one of tetraploid species. Yao et al. [3] and Chen et al. [23] also recognized cytologically that the B genome of T. petropavlovskyi was different from those of hexaploid wheats.

Relationships between T. petropavlovskyi and Ae. tauschii
Based on RFLPs, Ward et al. [5] found that T. petropavlovskyi is genetically more closely related to accessions of Ae. tauschii from Iran than from China. Yang et al. [10] concluded that T. petropavlovskyi is derived from a hybrid between Ae. tauschii and a presumed T. polonicum genotype. However, the phylogenetic analysis of the Acc-1 genes indicates that the D genome of T. petropavlovskyi is very similar to the D genome orthologs of T. aestivum and only distantly related to Ae. tauschii [26]. In the present study, D genomes of Ae. tauschii belong to two different clusters. One groups with SHW-DPW, T. compactum and T. aestivum cv. Chuannong-16, while T. petropavlovskyi, T. macha, T. spelta, T. sphaerococcum and three Chinese endemic wheats are included in a clade with Ae. tauschii (TA1691). Based on TCS analysis, T. petropavlovskyi shares common topologies with the hexaploid species D genomes. The sequence of the Pgk-1 gene from TA1691 is significantly different from those of other accessions of Ae. tauschii, in agreement with Huang et al. [31]. Together, available results supports that the D genome of T. petropavlovskyi is similar to D genome orthologs of T. aestivum and only distantly related to Ae. tauschii.

The divergence time of the T. petropavlovskyi
Huang et al. [31] report that the diploid progenitors of the A, B and D genomes present in diploid, tetraploid, and hexaploid wheats radiated between 2.5 and 4.5 MYA. We report that the divergence time of the A, B and D genomes corresponds to 2.61 (95% C.I., 1.77-3.55), 2.27 (95% C.I., 1.69-3.19) and 2.05 MYA (95% C.I., 1.40-2.81), respectively. The divergence of the B genome of T. petropavlovskyi from those of other wheats is dated in this paper is from 0.16 (95% C.I., 0-0.38) to 0.69 MYA (95% C.I., 0.55-0.70), the earliest date for the four Chinese endemic wheat landraces we considered.
Concerning the A and D genome of T. petropavlovskyi, the resulting divergence times are 0.14-0.33 and 0.11-0.27 MYA, respectively, values similar to those of other hexaploid species. The divergence time of A genome of T. petropavlovskyi from T. polonicum varies from 0.68 (95% C.I., 0.41-0.71) to 0.90 MYA (95% C.I., 0.45-1.41), a divergence earlier than those between T. petropavlovskyi and hexaploid species. The divergence time results indicate that T. polonicum may have played a role in the evolutionary history of T. petropavlovskyi.

The possible origin of T. petropavlovskyi
This study shows that the Pgk-1 sequences of the A genome of T. petropavlovskyi group with T. polonicum. For the Pgk-1 locus of the D genome, the accessions of Ae. tauschii just cluster with the   Table 1. Haplotypes in the network are represented by circles. Distance between nodes is proportional to the number of nucleotide substitutions among sequences. doi:10.1371/journal.pone.0071139.g005 Figure 6. TCS network inferred from the exon sequences of the Pgk-1 gene of T. petropavlovskyi and its related species. Abbreviations of species names are listed in Table 1. Haplotypes in the network are represented by circles of different color corresponding to the genomes indicated. doi:10.1371/journal.pone.0071139.g006 Evolutionary History of Triticum petropavlovskyi amphiploid SHW-DPW. The Pgk-1 B genome data indicate that T. petropavlovskyi is distantly related to the other three Chinese endemic wheat landraces. The MJ network results are congruent with the results reported above. Also the TCS analysis supports the conclusion that the relationship among haplotypes of T. petropavlovskyi and T. polonicum have very similar A and B genomes. We report a distant relationship between T. petropavlovskyi and Ae. tauschii. We conclude that T. petropavlovskyi is neither derived from an independent allopolyploidization event nor from a single mutation in T. aestivum. It is most likely that T. petropavlovskyi has an origin starting with a natural cross between T. aestivum and T. polonicum, with that event taking place around 0.78 MYA.