Phylogenetic relationships in Elymus (Poaceae: Triticeae) based on the nuclear ribosomal internal transcribed spacer and chloroplast trnL-F sequences

Authors


Author for correspondence: Bao-Rong Lu Tel./Fax: +86 21 65643668 Email: brlu@fudan.edu.cn

Summary

  • • To estimate the phylogenetic relationship of polyploid Elymus in Triticeae, nuclear ribosomal internal transcribed spacer (ITS) and chloroplast trnL-F sequences of 45 Elymus accessions containing various genomes were analysed with those of five Pseudoroegneria (St), two Hordeum (H), three Agropyron (P) and two Australopyrum (W) accessions.
  • • The ITS sequences revealed a close phylogenetic relationship between the polyploid Elymus and species from the other genera. The ITS and trnL-F trees indicated considerable differentiation of the StY genome species.
  • • The trnL-F sequences revealed an especially close relationship of Pseudoroegneria to all Elymus species included. Both the ITS and trnL-F trees suggested multiple origins and recurrent hybridization of Elymus species.
  • • The results suggested that: the St, H, P, and W genomes in polyploid Elymus were donated by Pseudoroegneria, Hordeum, Agropyron and Australopyrum, respectively, and the St and Y genomes may have originated from the same ancestor; Pseudoroegneria was the maternal donor of the polyploid Elymus; and some Elymus species showed multiple origin and experienced recurrent hybridization.

Introduction

Polyploidy, resulting from either duplication of a single but complete genome (autopolyploidy) or from combination of two or more differentiated genomes (allopolyploidy), is a prominent mode of speciation (Stebbins, 1971; Masterson, 1994; Soltis & Soltis, 2000). About 70% of angiosperms are of polyploid origin (Masterson, 1994; Soltis & Soltis, 2000; Wendel, 2000), which has significantly enriched diversity of plants. A better understanding of the processes of polyploidization and rapid diversification of the descendants of a single polyploidization event is therefore of widespread evolutionary interest (Wendel, 2000; Soltis et al., 2003). Recent studies using genetic markers in many genera suggest that recurrent origins for polyploid species are the rule rather than the exception (Soltis & Soltis, 2000), and that genetic diversity within recent polyploids is adequate to support rapid adaptive evolution (Doyle et al., 2003; Soltis et al., 2003).

The wheat tribe (Poaceae: Triticeae), an important gene pool for genetic improvement of cereal crops (Dewey, 1984; Lu, 1993, 1994), includes many autopolyploid and allopolyploid taxa. Data from extensive cytogenetic analyses have been used to illustrate systematic relationships of the tribe and to clarify the ancestry of many polyploid species. One complex group of polyploids within Triticeae is the genus Elymus that, following the taxonomic delimitation by Löve (1984) based essentially on genomic constitutions, includes approx. 150 perennial species distributed in a wide range of ecological habitats over the temperate and subtropic regions. Elymus has its origin through a typical allopolyploidy process (Dewey, 1984; Löve, 1984). Cytological studies suggest that five basic genomes, namely, the St, Y, H, P and W in various combinations constitute Elymus species (Lu, 1994). The St genome is a fundamental genome that exists in all Elymus species and is donated by the genus Pseudoroegneria (Dewey, 1967). The H, P and W genomes are derived from the genera Hordeum, Agropyron and Australopyrum of Triticeae, respectively (Dewey, 1971; Jensen, 1990; Torabinejad & Mueller, 1993). However, the donor of the Y genome that is present in the majority of the Asiatic Elymus species has not yet been identified, although extensive investigations have been carried out (Lu, 1993, 1994).

One important group of Elymus includes tetraploids with the StY genomes. About 30 StY genome Elymus species are found restrictedly in the temperate Asia, where more than half of the known Elymus species originated (Salomon & Lu, 1992; Lu, 1994). A large data set from cytological analyses of artificial hybrids among the StY genome Elymus species clearly indicates that the degree of chromosome pairing in the hybrids gradually decreases with increase in geographical distance from the locality of their parental species (Lu & Salomon, 1992; Lu, 1993). This means that the StY genomes in tetraploid Elymus species have been modified to a large extent and have relatively high genetic diversity. This phenomenon has not been found in the StH genomes, which have relatively high homology among different tetraploid Elymus species distributed in Asia (Lu et al., 1992). Knowledge of the molecular phylogeny of the StY genome Elymus will provide a better understanding of their genetic differentiation.

Molecular phylogenetic studies have successfully revealed the origins and evolutionary history of polyploids in plants, clarified the nature of different polyploids, and identified their parental lineages and the hybridization events involved in their formation (Soltis & Soltis, 1993; Wendel, 2000; Soltis et al., 2003). Comparative phylogenies between nuclear and chloroplast/mitochondrial sequences have become a powerful tool to identify the mode of polyploidization in particular groups (Ge et al., 1999; Mason-Gamer, 2001; Popp & Oxelman, 2001; Mason-Gamer, 2004; Rauscher et al., 2004). Among the available nuclear sequences, internal transcribed spacer (ITS) sequences have been used successfully in studying phylogenetic and genomic relationships of plants at lower taxonomic levels (Baldwin et al., 1995; Hsiao et al., 1995; Wendel et al., 1995; Zhang et al., 2002; Hao et al., 2004). The chloroplast DNA (cpDNA) sequences, particularly the noncoding regions such as the intron of trnL (UAA) and the intergenic spacer of trnL (UAA)–trnF (GAA) are also valuable source of markers for identifying the maternal donors of polyploids with additional capacity to reveal phylogenetic relationships of related species (Sang et al., 1997; Mason-Gamer et al., 2002; Xu & Ban, 2004).

In this study, we sequenced and analysed the nuclear ribosomal ITS and chloroplast trnL-F fragments for 30 Elymus polyploids and their putative diploid donors to explore the origin and relationships of the polyploid Elymus species. The objectives of this study were (1) to reveal relationships of the St, Y, P, W and H genomes of Elymus in relation to their putative diploid ancestors; (2) to elucidate the phylogenetic relationships of the StY genome Asiatic tetraploids; and (3) to determine the maternal genomic donor of Elymus polyploids.

Materials and Methods

Plant materials

A total of 57 Triticeae accessions were used in this study, including 45 Elymus accessions with different genomic combinations (i.e. the StY, StStY, StH, StPY, StWY and StHY genomes), five species of the related genus Pseudoroegneria (St and St1St2 genomes), three species of Agropyron (P genome), two species of Hordeum (H genome) and two species of Australopyrum (W genome). Bromus catharticus was used as the outgroup based on previous phylogenetic studies of Poaceae (Hsiao et al., 1995; Gaut, 2002). All seed materials were collected from the field or provided by Dr B. Salomon of the Swedish University of Agricultural Sciences, Sweden, and Drs H. Y. Zhou and H. Q. Zhang of Sichuan Agriculture University, China. All the accessions sequenced in this study, with their scientific names, geographic origins and GenBank accession numbers are listed in Table 1. The voucher specimens of this study are deposited in the Swedish University of Agricultural Sciences, Sweden and Sichuan Agriculture University, China.

Table 1.  Species of Elymus and other closely related genera used in this study
No.SpeciesAccession No.GenomeOriginGenBank Accession No.
ITStrnL-F
  1. Numbers before species’ names serve as identifiers of specific accessions that correspond to the numbers in Figs 1,2, and 3. GenBank accessions with an asterisk (*) represent previously published sequences from the GenBank (http://www.ncbi.nlm.nih.gov).

Pseudoroegneria (Nevski) A. Löve
1P. spicataPI547161StOregon, USAAY740793AF519159*
2P. libanoticaPI228389StIranAY740794AY730567
3P. strigosaPI499637StUrumqi, Xinjiang, ChinaAY740795AF519155*
4P. alashanicaZ2006St1St2Yinchuan, Ninxia, ChinaAY740796, AY740797AY73069
5P. elytrigioidesZ2005St1St2Changdu, Tibet, ChinaAY740798, AY740799AY730568
Hordeum L.
6H. bogdaniiPI531761HChinaAY740876AY740789
7H. brevisubulatumY1604HFuyun, Xinjiang, ChinaAY740877AY740790
Agropyron Gaertner
8A. cristatumH10154PAltai, Xinjiang, ChinaAY740890AY740791
9A. cristatumH10066PAltai, Xinjiang, ChinaAY740891AY740792
10A. mongolicumPL36482*AF519117*
Australopyrum (Tsvelev) A. Löve
11A. retrofractumCrane 86146WAF519118*
12A. pectinatumWL36483*, L36484*
Elymus L.
13E. antiquusH7087StYLixian, Sichuan, ChinaAY740814, AY740815
14E. antiquusH3400StYSichuan, ChinaAY740818, AY740819AY730581
15E. anthosachnoidesY2236StYYajiang, Sichuan, ChinaAY740820, AY740821AY740770
16E. barbicallusH3267StYChina, seeds from D. R. Dewey, 1988, D 2509AY740824, AY740825
17E. barbicallusH3268StYChina, seeds from D. R. Dewey, 1988, D 3512AY740822, AY740823AY730580
18E. brevipesY2245StYYajiang, Sichuan, ChinaAY740826, AY740827AY740771
19E. burchan-buddaeY3049StYHongyuan, Sichuan, ChinaAY740872, AY740873
20E. burchan-buddaeY2219StYBatang, Sichuan, ChinaAY740870, AY740871AY740772
21E. burchan-buddaeY2207StYZogang, Tibet, ChinaAY740874, AY740875
22E. ciliarisH7000StYBeijing, ChinaAY740830, AY740831AY740773
23E. dolichatherusH8024StYZhaojie, Sichuan, ChinaAY740834, AY740835
24E. dolichatherusY1411StYWenchuan, Sichuan, ChinaAY740836, AY740837AY730574
25E. grandisH3879StYLingtong, Shaanxi, ChinaAY740828, AY740829AY730572
26E. nakaiiH7386StYMaowen, Sichuan, ChinaAY740856, AY740857
27E. nakaiiH7371StYLi xian, Sichuan, ChinaAY740854, AY740855AY730585
28E. pendulinusH8986StYChangdu, Tibet, ChinaAY740846, AY740847AY730582
29E. pendulinusY1412StYWenchuan, Sichuan, ChinaAY740848, AY740849
30E. shandongensisH3202StYWuhan, Hubei, ChinaAY740816, AY740817AY730583
31E. tibeticusH8927StYLuhuo, Sichuan, ChinaAY740862, AY740863
32E. tibeticusH8366StYGongbogyamda, Tibet, ChinaAY740864, AY740865
33E. aboliniiH3306StYAlma-Ata, Medeo, KazakstanAY740899, AY740900
34E. aboliniiH8491StYXinyuan, Xinjiang, ChinaAY740812, AY740813AY730584
35E. canaliculatusH4116StYNWFP, PakistanAY740832, AY740833AY740774
36E. fedtschenkoiH7510StYHabahe, Xinjiang, ChinaAY740838, AY740839AY740775
37E. fedtschenkoiH4114aStYNWFP, PakistanAY740840, AY740841
38E. glaberrimusY2042StYJeminay, Xinjiang, ChinaAY740844, AY740845AY740777
39E. gmeliniiH1033StYAltai, Xinjiang, ChinaAY740842, AY740843AY740776
40E. longearistatusH3261StYNorthern IranAY740804, AY740805AY740778
41E. longearistatusH4114bStYNWFP, PakistanAY740806, AY740807AY740779
42E. macrochaetusH10303StYGissar, TadzhikistanAY740850, AY740851AY730573
43E. macrochaetusH10208StYGissar, TadzhikistanAY740852, AY740853
44E. nevskiiH3305StYChatkal, UzbekistanAY740860, AY740861AY740780
45E. nevskiiH10213StYGissar, TadzhikistanAY740858, AY740859AY730571
46E. semicostatusH3288StYMandi, PakistanAY740800, AY740801AY730576
47E. semicostatusH4101StYHazara, PakistanAY740802, AY740803AY730575
48E. validusH4078StYBabusar, Gilgit, PakistanAY740866, AY740867AY730578
49E. validusH4100StYNaran village, Hazara, PakistanAY740868, AY740869AY730579
50E. caucasicusH3207StYDilidjan, ArmeniaAY740808, AY740809AY730577
51E. panormitanusH4152StYCrimea, UkraineAY740810, AY740811AY730570
52E. tschimganicusH3302StStYXinjiang, ChinaAY740878, AY740879,AY740880AY740783
53E. caninusPI564910StHRussian FederationAY740897, AY740898AY740781
54E. himalayanusH4134StHYAstor valley, Gilgit, PakistanAY740881, AY740882,AY740883AY740782
55E. rectisetusH3152StWYLake Lyndon, AustraliaAY740893, AY740894
AY740896, AY740786
AY740895
56E. melantherusZY3146StPYSichuan, ChinaAY740884, AY740885,AY740886AY740785
57E. rigidulusZY3113StPYXiahe, Gansu, ChinaAY740887, AY740888,
AY740889
AY740784
Bromus L.
58B. catharticusS20004Kunming, Yunnan, ChinaAF521898*AY829228

DNA extraction and purification

Seeds were germinated and grown in a growth chamber with 15 h in light and 9 h in dark at 25°C. Leaf samples collected from each accession at the seedling stage were ground in liquid nitrogen in a 1.5-ml microfuge tube, and DNA was extracted and purified using a slight modification of the cetyltrimethylammonium bromide (CTAB) procedure outlined in Doyle and Doyle (1990).

ITS amplification, cloning, and sequencing

The internal transcribed spacer (ITS) region of the nuclear ribosomal DNA was amplified by polymerase chain reaction (PCR) using the primers of ITS4 and ITS5 (Hsiao et al., 1995). The PCR amplification of ITS DNA was carried out in a total reaction volume of 25 µl containing 1× reaction buffer, 1.5 mm MgCl2, 0.5 µm of each primer, 200 µm of each dNTP (TaKaRa Inc., Dalian, Liaoning, China), 0.5 units of ExTaq Polymerase (TaKaRa Inc.), with an addition of 8% dimethyl sulfoxide (DMSO) and water to the final volume. The thermocycling profile consisted of an initial denaturation step at 94°C for 3 min, followed by 35 cycles of 0.5 min at 94°C, 1 min at 55°C, 1 min at 72°C and final extension step of 10 min at 72°C. PCR reactions from the polyploid Elymus were run in triplicates in different thermocyclers and the PCR products were combined in an attempt to offset the potential effects of PCR drifts (Wagner et al., 1994). The PCR products were purified using a gel extraction kit (TaKaRa Biotechnology (Dalian) Co., Ltd, Dalian, China) and linked into a pMD-T vector according to the manufacturer's instruction (TaKaRa Biotechnology (Dalian) Co., Ltd). Transformation, plating and isolation of plasmids were performed as described in Sambrook et al. (1989). Purified plasmid DNAs were digested with EcoRI and HindIII. For each of the Elymus species, 10–15 cloned PCR products were sequenced to include all the possible ITS sequences from the donor species, using ABI BigDye terminators according to the product instructions, and run on an ABI 3730 sequencer.

CpDNA amplification and sequencing

The chloroplast tRNA genes trnT, trnL-5′, trnL-3′ and trnF, along with their intervening noncoding regions, were amplified using the primers a and f, or c and f of Taberlet et al. (1991). Amplification of the cpDNA was performed in a total reaction volume of 25 µl with the same components as described in the ITS amplification, except that DMSO was not added. The PCRs were performed as for the ITS nuclear ribosomal DNA, except the annealing and extension times were 1.5 min each. The PCR products were cleaned as described in the previous section, the primers b, d, e and f of Taberlet et al. (1991) were used to sequence both strands of the PCR fragments to unambiguously identify all sites, and the sequencing reactions were as described earlier.

Phylogenetic analysis

The ITS and trnL-F sequences were aligned with clustal x (Thompson et al., 1999) and refined manually. The boundaries of the ITS region (ITS1–5.8S–ITS2) and trnL-F (trnL intron–trnL 3′exon–intergenic spacer–trnF 5′exon) were determined according to Hsiao et al. (1995) and Ogihara et al. (2002), respectively. Gaps were coded as binary characters by their presence/absence, and were used for the phylogenetic analyses. The basic sequence statistics, including nucleotide frequencies, transition/transversion (ns : nv) ratio and variability in different regions of the sequences were computed by mega 3 (Kumar et al., 2004).

The aligned sequences were used as the input data for paup 4.0 (Swofford, 1998), mega 3 and the phylip software package (Felsenstein, 1995). Parsimony analyses were performed by heuristic search with tree bisection-reconnection (TBR) branch swapping, MULPARS option, ACCTRAN optimization, and 100 random addition replicates. Topological robustness was assessed by bootstrap analysis with 500 replicates using simple taxon addition. The aligned sequences were also analyzed with the Neighbor-Joining Program of the phylip package, and carried out with 1000 bootstrap replicates.

Results

Variation in ITS and trnL-F sequences

The ITS sequences in this study included three regions: (1) 43 nucleotides of the 18S rRNA gene; (2) the complete sequences of ITS1, 5.8S rRNA gene and ITS2; and (3) 58 nucleotides of the 26S rRNA. Sequences of the 18S rRNA and the 26S rRNA genes showed no variation for all accessions included in this study. The length of sequences ranged from 212 to 221 bp in the ITS1 region, and from 216 to 218 bp in the ITS2 region. The 5.8S rRNA gene was 164 bp long and completely identical for all the cloned sequences of the 58 accessions. The average of G + C content was 61.2%. Sequence alignment necessitated gaps of one to eight bases in length. Of 223 variable sites, 115 were parsimoniously informative.

The trnL-F fragment sequenced in this study included four regions: (1) the partial trnL intron; (2) the trnL 3′ exon; (3) the trnLtrnF intergenic spacer; and (4) the partial trnF exon with 40 bp. The length of the sequenced chloroplast trnL-F varied from 859 to 882 bp in all accessions. The average of G + C content was 30.2%. Of the 210 variable sites, 107 were parsimoniously informative, including the polymorphisms introduced by insertions/deletions.

Phylogenetic analysis of the nuclear ITS sequences

We first constructed the phylogeny of all the diploid species based on their ITS sequences in order to reveal relationships of the putative genomic donors of Elymus. Three equally most parsimonious trees were obtained with the tree length of 147 steps, a consistency index (CI) of 0.857 and a retention index (RI) of 0.764. As shown in the strict consensus tree (Fig. 1), species with the same genomes formed highly supported monophylies, indicating that the four genomes have been well differentiated and corresponded to different morphologically recognized genera. The H genome (Hordeum) was at the basal position, followed by the St genome (Pseudoroegneria) that was sister to the other two related P and W genomes (Agropyron and Australopyrum).

Figure 1.

The strict consensus tree of three most parsimonious (MP) trees inferred from the ITS sequences of the diploid species (Tree length = 147, consistency index (CI) = 0.857, retention index (RI) = 0.764). The topologies obtained by Neighbor-Joining method (NJ) are the same except for some nodes having different bootstrap values. Numbers above and below the branches indicate bootstrap values > 50% by MP and NJ analyses, respectively. Numbers after the species names refer to accession numbers as indicated in Table 1. Capital letters in parentheses following the species names indicate the genome type of the species. The genome type (St, P, W or H) of a monophyletic group is given to the right.

To further analyse genomic relationships and the origin of the polyploid Elymus, ITS sequences of all the polyploid species (Elymus polyploids with the StH, StY, StStY, StHY, StPY and StWY genomes and Pseudoroegneria tetraploids with the St1St2 genomes) were included in the phylogenetic analysis, together with those of diploids containing the St, H, P and W genomes. To ensure obtaining all possible types of ITS sequences of a polyploid, 10–15 clones from each of the selected Elymus species were sequenced. In the case of multiple identical sequences resulting from cloned PCR products of one accession, only one sequence was included in the data set. Consequently, 107 unique sequences were obtained and used for phylogenetic analyses.

Maximum parsimony analysis resulted in 440 equally most parsimonious trees. Each of the trees was 431 steps with a CI of 0.620 and a RI of 0.819. In one of the most parsimonious tree (Fig. 2), all the homeologous ITS sequences from polyploid accessions grouped with those of the diploid parental clades expected from cytological studies. Four major clades with high bootstrap support (83–100%) were found, which correspond to the four genomic types (H, P, W and St). The first clade consisted of the Hordeum and Elymus species with the StH and StHY genomes (100% bootstrap support). The second clade included the Australopyrum and StWY genome Elymus species (96% bootstrap support), while the third clade included the Agropyron and StPY genome Elymus species (95% bootstrap support). The fourth (largest) clade (83% bootstrap support) comprised the Pseudoroegneria species (St and St1St2 genomes) and all polyploid Elymus with the StStY, StY, StH, StHY, StPY, and StWY genomes. Three subclades (A, B, and C subclades) were recognized in this clade. It is worth mentioning that no obvious Y-genome specific clade was detected in the phylogenetic tree. Neighbor-Joining analysis generated a similar topology with minor variation in bootstrap values.

Figure 2.

One of the 440 most parsimonious (MP) trees inferred from the internal transcribed spacer (ITS) sequences of all the accessions used in this study (Tree length = 431, consistency index (CI) = 0.620, retention index (RI) = 0.819). The topologies obtained by Neighbor-Joining method (NJ) are the same except for some nodes having different bootstrap values. Numbers above and below the branches indicate bootstrap values > 50% by MP and NJ analyses, respectively. Branch lengths are proportional to the number of nucleotide substitutions; the scale bar at the upper-left corner indicates one substitution. Numbers after species names refer to the accession numbers shown in Table 1. Capital letters in parentheses indicate the genome type of the species. Names in boldface indicate species of Pseudoroegneria and other diploid genera. The arrow indicates the 8-bp deletion shared by species in the subclade. *, polyploid species with the ‘H’ genome; **, polyploid species with the ‘W’ genome; ***, polyploid species with the ‘P’ genome. The genome type (St, P, W or H) of a monophyletic group is given to the right.

For clarity, we named the four clades using genomic symbols of the diploid species, i.e. the H, W, P, and St clades, respectively (Fig. 2). The phylogenetic tree showed that all Elymus polyploids, except for those with the St and Y genomes, had two distinct types of ITS sequences with one forming a clade with its respective putative diploid donors and the other grouped with the St genome clade. By contrast, all Elymus species with the Y and St genomes (StY and StStY) were retained in the St clade. In addition, a well-supported branch (97% bootstrap) was found within the St clade, including the St1St2 genome Pseudoroegneria and the StY genome Elymus tetraploids that shared an 8-bp deletion in their ITS sequences (Fig. 2).

Phylogenetic analysis of the chloroplast trnL-F sequences

The chloroplast trnL-F sequences of all Elymus polyploids and their putative diploid donor species were included for phylogenetic analysis. Maximum parsimony analysis resulted in 316 equally most parsimonious trees with 307 steps, a consistency index of 0.785, and a retention index of 0.816. It was evident from the phylogenetic tree (Fig. 3) that the diploid species grouped into four distinct clades corresponding to the St, P, W and H genomes, respectively. The tetraploid Pseudoroegneria (St1St2) and all Elymus species formed a large but highly supported clade (100%) together with the diploid Pseudoroegneria species, suggesting a close relationship between Pseudoroegneria and Elymus species in terms of their chloroplast genomes. This clade was named as St clade because all diploid and polyploid accessions contained the St genome. The diploid Hordeum species (H clade) was the earliest divergent lineage, followed by the diploid Agropyron (P clade) and Australopyrum (W clade) species that were sisters to the St clade. Neighbor-Joining analysis produced a similar topology with only very minor differences in bootstrap supports (Fig. 3).

Figure 3.

One of 316 most parsimonious (MP) trees inferred from the trnL-F sequences of all the accessions used in this study (Tree length = 307, consistency index (CI) = 0.785, retention index (RI) = 0.816). The topologies obtained by Neighbor-Joining method (NJ) are the same except for some nodes having different bootstrap values. Numbers above and below the branches indicate bootstrap values greater than 50% by MP and NJ analyses, respectively. Branch lengths are proportional to the number of nucleotide substitutions, and the scale bar at the upper-left corner indicates one substitution. Capital letters in parentheses indicate the genome type of the species. Names in boldface indicate species of Pseudoroegneria and other diploid genera. *, polyploid species with the ‘H’ genome; **, polyploid species with the ‘W’ genome; ***, polyploid species with the ‘P’ genome. The genome type (St, P, W or H) of a monophyletic group is given to the right.

Discussion

Phylogenetic relationships of Elymus and its proposed diploid ancestors

The genus Elymus consists of polyploids that are widely distributed over different continents and includes a large number of endemic species. Only a few molecular studies addressing phylogenetic relationships of the StH and StHY genome Elymus species are reported (Mason-Gamer, 2001; Mason-Gamer et al., 2002; McMillan & Sun, 2004; Xu & Ban, 2004). Little is known about phylogeny of the Asiatic StY genome Elymus species at molecular level. Analyses of ITS sequences collected from a wide range of polyploid Elymus species and their related genera will provide opportunities for understanding their phylogenetic relationships, ancestral donors and polyploidization events in the speciation processes.

In the diploid and polyploid ITS trees, four of the five genomes presented in Elymus formed distinct clades, the exception being the Y genome. Of the four clades, the H genome clade was basal, with the W, P, and St clades being successively more distantly related. There was no obvious Y genome clade, and all the StY species were placed in the St genome clade. These results indicate that ITS sequences of all the genomes derived from the diploid ancestors have remained clearly differentiated in the polyploid Elymus. This can be reflected by the fact that all the allopolyploid species (except for those with the Y genome) contained two distinct types of ITS sequences, with one type in the St clade and the others in different genomic clades (H, P or W clade), respectively. This strongly suggests that ITS sequences in different Elymus species showed a clear linkage with those in their diploid ancestors. This is illustrated by the fact that the StH, StHY, StPY and StWY genome Elymus species were simultaneously clustered in both of their ancestral groups, indicating that two distinct types of ITS sequences exist in these polyploid Elymus. Obviously, the homogenization of ITS sequences is not significant in the polyploid Elymus. This provides strong evidence that the polyploid Elymus species are derived from polyploidization through hybridization between different ancestral genera, as indicated by cytological analyses (Dewey, 1984; Lu, 1994). For example, the StH genome Elymus species are derived from the hybridization between Pseudoroegneria (St) and Hordeum (H). Although the rDNA sequences are considered to undergo a rapid homogenization through concerted evolution, growing evidence shows that ITS polymorphism or incomplete homogenization is the rule rather than the exception (Buckler et al., 1997; Hershkovitz et al., 1999). Directional and bidirectional interlocus concerted evolution following allopolyploid speciation have been documented (Wendel et al., 1995; Fulnecek et al., 2002), but allopolyploid species often maintain both parental sequences of the ITS region. The phenomenon of incomplete homogenization is useful for understanding the origin of hybrids and genomic constructions in polyploidy species (Ainouche & Bayer, 1997; Hershkovitz et al., 1999; Popp & Oxelman, 2001; Yonemori et al., 2002; Koch et al., 2003). It is evident that the homogenization of ITS sequences in the allopolyploid Elymus is not completed, which probably suggests the recent origin of these polyploids because sufficient time is needed to allow ITS sequences to be homogenized (Ainouche & Bayer, 1997).

By contrast, all the tetraploid StY genome Elymus species had only one genomic type of ITS sequences since all these species were included in the St clade (Fig. 2). This was also true of the tetraploid Pseudoroegneria species. However, the hexaploids that combined the St and Y genomes with one of the three differentiated genomes (H, P or W genome) had two genomic types of ITS sequences. There are two possible explanations for this phenomenon. One of the explanations is that the St and Y genomes may have the same origin, because the St genome clade in this study did not show any St- and Y-genome-associated subclades. This explanation is supported by the relatively close affinities between the St and Y genomes reported by Lu et al. (1992), and Lu & Bothmer (1991) based on cytological investigations, although a close affinity between the St and P genomes are also reported (Wang et al., 1985). In addition, no diploid Y genome species has been found so far despite great efforts world-wide, further supporting the this explanation. It is possible that homogenization of ITS sequences has occurred between the St and Y genomes in polyploid Elymus species. However, since there is no (or extremely low) chromosome pairing between the St and Y genomes under normal conditions (the situation is the same between the St and H, P or W genomes) the likelihood of homogenization of the ITS sequences between the St and Y genomes seems to be low in the polyploid Elymus species. Previous studies also indicate that homogenization of ITS sequences in allopolyploids rarely results in the formation of a single genomic type of sequences (Wendel et al., 1995; Rauscher et al., 2004). Therefore, we prefer the explanation that the St and Y genomes may have the same origin.

Differentiation of the StY genome Elymus tetraploids

Previous cytological investigations suggest that the StY genomes in Elymus tetraploids are considerably differentiated in relation to their geographical distribution (Lu & Salomon, 1992; Lu, 1993). The analysis of ITS sequences based on nearly all the StY genome Elymus species covering a wide range of distribution also demonstrated genomic differentiation of these species. It is obvious that a subclade including two Pseudoroegneria (St1St2) tetraploids and eight Elymus (StY) species are supported with high bootstrap support (97%) and are differentiated from other StY species (Fig. 2). This provides an evidence of genomic differentiation of the StY genome species from molecular data. Genomic differentiation among species through polyploidy or at the same ploidy level usually results in great diversity of polyploid species, as suggested by Soltis & Soltis (2000). However, because of the weak resolution of the current ITS data in the St clade, whether or not the differentiation of the StY genome Elymus species is associated with their geographic distribution, as suggested by previous cytological observations, needs to be verified further with additional evidence.

Multiple origin is a very important evolutionary process of polyploid species, which emphasizes ‘each polyploid species forms over and over again from different parental genotypes generating a diverse array of polyploid genotypes’ (Soltis & Soltis, 1999). The ITS tree in this study also suggested a multiple origin of some StY genome species resulting from recurrent hybridization, which can be shown by different accessions of the same species, will appear at different clades of a phylogenetic tree. For example, in this study, different accessions of Elymus antiquus (−13 and −14) and Elymus pendulinus (−28 and −29) were grouped in different subclades of the St clade. This helps to explain the abundant genetic diversity within an Elymus species. The recurrent hybridization also promoted rapid adaptation of the Elymus species to different ecological habitats, resulting in the formation of many endemic genotypes and species.

It is worth pointing out that the two Pseudoroegneria and eight StY genome Elymus species from the central and eastern Asia consistently had an 8-bp deletion in their ITS sequences and clustered distinctly into one subclade. It is most likely that the 8-bp deletion had already occurred in the Pseudoroegneria species before being passed on to some Elymus species during the polyploidization process. If this assumption holds true, the St genome in Elymus should have been derived from more than one Pseudoroegneria species/populations through hybridization. In other words, Elymus species with the 8-bp deletion in the ITS sequences may have evolved from hybridization of the Pseudoroegneria ancestor with the 8-bp deletion, whereas other Elymus species without the 8-bp deletion might originate from hybridization of other Pseudoroegneria ancestors. This could also explain the rich diversity and wide adaptation of the StY genome Elymus species (Lu, 1994).

The maternal donor of Elymus species

The trnL-F gene tree represents a maternal genealogy of the Elymus species, because the chloroplast genome is maternally inherited in grasses (Ge et al., 1999; Mason-Gamer et al., 2002). This offers an opportunity to identify the maternal parents of the Elymus species. All polyploid species including the Pseudoroegneria and Elymus species formed a highly supported monophyletic group (100% and 92% bootstraps for MP and NJ trees, respectively) on the maternal trnL-F tree. This suggests that the Pseudoroegneria species (St genome) served as the maternal donor during the polyploid speciation of the Elymus species (tetraploids and hexaploids). This result, in conjunction with the biparentally inherited ITS tree (Fig. 2), implies that diploid Hordeum (H genome), Agropyron (P) and Australopyrum (W) species were the paternal parents for the Elymus species with the H, P and W genomes. This conclusion is in good agreement with previous studies by McMillan & Sun (2004) and Mason-Gamer et al. (2002) using restriction fragment length polymorphism (RFLP) analysis of cpDNA and chloroplast DNA sequence data, where Pseudoroegneria was suggested as the chloroplast genome donor of the northern American StH genome Elymus and two StY genome species. A recent study based on analysis of partial trnL-F sequences of a few Elymus species by Xu & Ban (2004) also presented similar results.

The phylogenetic tree based on trnL-F sequences of the Elymus species also indicated a multiple origin of polyploids in the evolutionary process of some Elymus species. For example, different accessions of Elymus longearistatus were clustered in the different subclades of the St clade. In addition, all Elymus species used in this study were scattered in different subclades in the trnL-F tree, which may also suggest different maternal lineages of the polyploid genus. Conversely, in the ITS tree, at least two hybridization events involving Elymus caninus (StH) and Elymus himalayanus (StHY) occurred to form these polyploids with the H genome species because they appeared in different subclades of the St clade (E. caninus in the A-subclade and E. himalayanus in the B-subclade). In addition, two hybridization events involving Elymus melantherus (StPY), Elymus rigidulus (StPY), and Elymus rectisetus (StWY) can be identified for generating the polyploid species with the P and W genomes. Therefore, nuclear DNA and cpDNA sequences showed that many Elymus species had a multiple origin and experienced recurrent hybridization between species with different genomes. This suggests that hybridization and polyploidization were the major driving force in the diversity and evolution of the genus Elymus.

Acknowledgements

This research was supported by the National Natural Science Foundation of China (Grant no. 30270092) and the Program for Key International S & T Cooperation Project of China (Grant no. 2001CB711103). Prof. M. E. Barkworth of the Utah State University, USA, provided critical comments. Dr N. Clarke of the Norwegian Forest Research Institute, Norway, kindly edited English of the manuscript.

Ancillary