Progenitor-derivative relationships of Hordeum polyploids (Poaceae, Triticeae) inferred from sequences of TOPO6, a nuclear low-copy gene region.

Polyploidization is a major mechanism of speciation in plants. Within the barley genus Hordeum, approximately half of the taxa are polyploids. While for diploid species a good hypothesis of phylogenetic relationships exists, there is little information available for the polyploids (4×, 6×) of Hordeum. Relationships among all 33 diploid and polyploid Hordeum species were analyzed with the low-copy nuclear marker region TOPO6 for 341 Hordeum individuals and eight outgroup species. PCR products were either directly sequenced or cloned and on average 12 clones per individual were included in phylogenetic analyses. In most diploid Hordeum species TOPO6 is probably a single-copy locus. Most sequences found in polyploid individuals phylogenetically cluster together with sequences derived from diploid species and thus allow the identification of parental taxa of polyploids. Four groups of sequences occurring only in polyploid taxa are interpreted as footprints of extinct diploid taxa, which contributed to allopolyploid evolution. Our analysis identifies three key species involved in the evolution of the American polyploids of the genus. (i) All but one of the American tetraploids have a TOPO6 copy originating from the Central Asian diploid H. roshevitzii, the second copy clustering with different American diploid species. (ii) All hexaploid species from the New World have a copy of an extinct close relative of H. californicum and (iii) possess the TOPO6 sequence pattern of tetraploid H. jubatum, each with an additional copy derived from different American diploids. Tetraploid H. bulbosum is an autopolyploid, while the assumed autopolyploid H. brevisubulatum (4×, 6×) was identified as allopolyploid throughout most of its distribution area. The use of a proof-reading DNA polymerase in PCR reduced the proportion of chimerical sequences in polyploids in comparison to Taq polymerase.

Introduction

Polyploidization, or whole genome duplication, is a major mechanism in plant evolution. Numerous studies have tried to evaluate the proportion of polyploidy in angiosperms, varying widely between 30 and 80% [1]. It is now acknowledged that probably all angiosperm lineages experienced one or several rounds of polyploidization in their history [2]–[4]. Otto and Whitton [5], analyzing the distribution of haploid chromosome numbers, estimated that polyploidy might be involved in about 2–4% of speciation events, thus proposing that polyploidization “may be the single most common mechanism of sympatric speciation in plants” ([5], p. 427).

Two different concepts for the definition of the type of polyploidy exist. On the one hand is the classic cytogenetic definition where the presence of only bivalent-forming chromosomes during meiosis characterizes allopolyploids while multivalent formation of homoeologous chromosomes indicates autopolyploidy [6], [7]. The second definition is based on a taxonomic concept, where polyploids formed through hybridization of different species (allopolyploids) contrast with hybrids formed through genome duplication or crossing of different genotypes from within a species (autopolyploids). Taxonomic allopolyploids are often termed segmental allopolyploids in the cytogenetic reference frame, indicating the presence of only locally differentiated chromosomes. We here use the taxonomic system of polyploid definition and explicitly refer to cytogenetic allopolyploids by indicating their genome composition.

The grass genus Hordeum L. belongs to the economically important tribe Triticeae and consists of 33 species (including cultivated barley, H. vulgare) distributed in arid and temperate regions of the world with two centers of diversity in Eurasia and in southern South America [8], [9]. Cytogenetic studies of meiotic chromosome behavior in interspecific hybrids led to the definition of four different genomes within Hordeum. For genome denomination we follow Blattner [9], with the H genome occurring in H. vulgare and H. bulbosum, Xu in H. murinum, Xa in H. marinum and H. gussoneanum, and I in the remaining species. The genus is particularly well suited to the study of polyploidization, as nearly half of the species are polyploids, with tetraploid (2n = 4x = 28) and hexaploid (2n = 6x = 42) taxa and cytotypes, comprising allo- and autopolyploids [8].

During the last 50 years numerous analyses have been carried out to study relationships in Hordeum [9] due to the potential of the wild relatives to improve barley. For the diploid taxa of Hordeum, datasets that include multiple nuclear loci converge to similar results [9]. Thus, a good phylogenetic hypothesis seems to be accomplished for this group, although studies including several individuals per species are yet rare. For polyploid species and cytotypes of Hordeum phylogenetic relationships were studied mainly for small taxon groups [10]–[13] using nuclear low-copy number loci. Blattner [14] conducted a thorough phylogenetic analysis of all taxa of the genus, including mostly multiple individuals per taxon. This analysis used nuclear rDNA internal transcribed spacer (ITS) sequences as molecular markers, which can undergo unidirectional homogenization [15] or loss of rDNA clusters [16] and thus are not ideal for the study of polyploid evolution. Therefore, a phylogenetic study of all Hordeum polyploids based on several individuals per species and a single- or low-copy nuclear locus is still lacking, which severely restricts evolutionary studies in these taxa.

In this study we report results from a phylogenetic analysis of all Hordeum species using cloned sequences of the nuclear low-copy region TOPO6 [12] that is a partial sequence of the Spo11 gene, a conserved plant homologue of the widespread archaean topoisomerase VI subunit A involved in inducing meiotic DNA double-strand breaks during recombination [17]–[18]. It consists of an unusually high number of introns [19] with exons conserved enough to design PCR primers. To determine genetic diversity of this locus within species and to be able to detect possible independent origins of polyploids we included for all taxa except one (H. guatemalense) an average of five individuals per taxon, representing the geographic distribution of the species. The aims of our study are to (i) define parental species of allopolyploids, (ii) analyze the status of putative autopolyploids, (iii) infer single or multiple origins for polyploids, (iv) infer speciation events on the polyploid level, (v) check for indications of introgression from outside of Hordeum, and (vi) compare the influence of two different types of DNA polymerases in PCR on the results of polyploid analyses. We do not assume that the gene tree obtained from the analyzed locus will represent the ‘true’ species phylogeny of the genus. This will, however, not impair the objectives of our study, as we are mainly interested in progenitor–derivative relationships between diploids and polyploids, which should be independent from the gross topology of the diploids in a phylogenetic tree.

Materials and Methods

Plant materials

We included 341 individuals representing all 33 species and most subspecies of the genus plus one individual of each of eight diploid Triticeae species outside Hordeum as outgroups (Table 1). Included individuals were obtained from germplasm repositories or sampled from natural populations (Table S1) and all necessary permits were acquired to use these materials. Herbarium vouchers of the analyzed materials were deposited in the herbaria of the IPK Gatersleben (GAT) or the Museum of Natural History, Buenos Aires (BA).

Table 1

Taxa included in the study.

Taxon	Ploidy level (N)1	Haploid genome	Native distribution area
Hordeum subgenus Hordeum
Section Hordeum
H. vulgare L.2
subsp. spontaneum (C.Koch.) Thell.	2× (2)	H	SW Asia
H. bulbosum L.	2× (5), 4× (3)	H, HH	Mediterranean to C Asia
Section Trichostachys Dum.
H. murinum L.
subsp. glaucum (Steud.) Tzvel.	2× (3)	Xu	Mediterranean to C Asia
subsp. murinum	4× (3)	XuXu	NW Europe to Caucasus
subsp. leporinum (Link) Arc.	4× (4), 6× (3)	XuXu, XuXuXu	Mediterranean to C Asia
Hordeum subgenus Hordeastrum (Doell) Rouy
Section Marina (Nevski) Jaaska
H. gussoneanum Parl.	2× (4), 4× (6)	Xa, XaXa	Mediterranean to C Asia
H. marinum Huds.	2× (3)	Xa	Mediterranean
Section Stenostachys Nevski
Series Sibirica Nevski
H. bogdanii Will.	2× (41)	I	C Asia
H. brevisubulatum (Trin.) Link2	2× (5), 4× (9), 6× (4)	I, II, III	C Asia
H. roshevitzii Bowden	2× (17)	I	C Asia
Series Critesion (Raf.) Blattner
H. californicum Covas & Stebb.	2× (13)	I	SW California
H. chilense Roem. & Schult.	2× (22)	I	Chile and W Argentina
H. comosum Presl	2× (24)	I	S Argentina
H. cordobense Bothmer et al.	2× (22)	I	C Argentina
H. erectifolium Bothmer et al.	2× (1)	I	C Argentina
H. euclaston Steud.	2× (14)	I	C Argentina, Uruguay
H. flexuosum Steud.	2× (8)	I	E+C Argentina
H. intercendens Nevski	2× (7)	I	SW California, NW Mexico
H. muticum Presl	2× (10)	I	C to N Andes
H. patagonicum (Haum.) Covas2	2× (14)	I	S Argentina
H. pubiflorum Hook.f.2	2× (17)	I	S Argentina
H. pusillum Nutt.	2× (13)	I	C+E USA
H. stenostachys Godr.	2× (19)	I	C Argentina
H. depressum (Scribn. & Sm.) Rydb.	4× (4)	II	W USA
Interserial allopolyploids of series Critesion and Sibirica
H. brachyantherum Nevski	4× (3)	II	W North America, Kamchatka, Newfoundland
H. fuegianum Bothmer et al.	4× (3)	II	S Argentina, S Chile
H. guatemalense Bothmer et al.	4× (1)	II	Guatemala, S Mexico
H. jubatum L.	4× (4)	II	NE Asia, NW+W North America, C Argentina
H. tetraploidum Covas	4× (4)	II	C Argentina
H. arizonicum Covas	6× (3)	III	SW USA
H. lechleri (Steud.) Schenk	6× (7)	III	C+S Argentina
H. parodii Covas	6× (4)	III	C Argentina
H. procerum Nevski	6× (4)	III	S Argentina
Section Nodosa (Nevski) Blattner
H. brachyantherum Nevski	6× (2)	IIXa	C California
H. capense Thunb.	4× (2)	IXa	S Africa
H. secalinum Schreb.	4× (4)	IXa	Mediterranean to W Europe
Outgroup species
Dasypyrum villosum (L) Candargy	2× (1)	V
Eremopyrum triticeum (Gaertn.) Nevski	2× (1)	FXe
Psathyrostachys juncea (Fisch.) Nevski	2× (1)	Ns
Taeniatherum caput-medusae (L.) Nevski	2× (1)	Ta
Triticum monococcum L.	2× (1)	Am
Triticum urartu Tumanian ex Gandilyan	2× (1)	Au
Secale strictum (C. Presl) C. Presl	2× (1)	R
Secale vavilovii Grossh.	2× (1)	R

Number of individuals included per species or cytotype;

species with subspecies not further detailed here.

Molecular methods

Genomic DNA was extracted from approximately 10 mg of silica gel-dried leaves with the DNeasy Plant Mini Kit (Qiagen) according to the protocol of the manufacturer. DNA quality and concentrations were checked on 1% agarose gels.

TOPO6 was amplified as described in Jakob and Blattner [12] using primers Top6-15F (5′-GTG YTG TST YCA ACT GAA GTC-3′) and Top6-17R (5′-CGT ACT CCA RYG CCA TTT C-3′) designed to bind in exons 15 and 17 of the gene. Thus the amplification products consist of introns 15 and 16 together with exon 16, and are of lengths between 800 and 1200 base pairs (bp) in many pooid grasses (Blattner, unpublished). PCR was conducted for all but 12 of the analyzed individuals, using 1 U of a standard DNA polymerase (Qiagen Taq DNA polymerase) in 50 µl reaction volume containing approximately 10–50 ng of genomic DNA, 1× Coral Load PCR Buffer (Qiagen), 1× Q Solution (Qiagen), 1.25 mM MgCl2, 0.2 mM of each dNTP, and 0.5 µM of each primer. The amplification process consisted of initial denaturation for 2 min at 95°C, followed by 35 cycles of 30 sec at 96°C, 1 min at 56°C, 2.3 min at 72°C, and a final extension of 12 min at 72°C. To reduce PCR errors, which become visible when amplicons are cloned, amplification of TOPO6 was performed in 27 polyploids and five recalcitrant diploid individuals using 1 U proof-reading polymerase (Finnzymes OY, Phusion Hot Start DNA polymerase) with the same PCR conditions as before but using the supplied 1× Phusion HF Buffer. Amplification conditions were modified as suggested by the provider with higher denaturation (98°C) and annealing temperatures (59°C).

Amplicons were purified using Nucleofast 96 Spin Plates (Macherey-Nagel) according to the protocol of the manufacturer, eluted in 20 µl of TE buffer, and sequenced on an ABI 3730XL automatic DNA sequencer (Applied Biosystems). For most of the diploid Hordeum and outgroup species amplicons were directly sequenced, while for all Hordeum polyploids and eight diploids amplicons were ligated into the pJET1.2 vector (Fermentas) and transformed into DH5α E. coli strains. On average 15 colonies per individual were randomly selected for screening the insertion of a TOPO6 fragment via PCR employing the primers pJET-F and pJET-R (Fermentas). Colonies showing products of the correct size (about 900 bp) were transferred to 200 µl LB broth medium with 0.1 mg/ml ampicilline and incubated overnight at 37°C. For a total of 952 colonies plasmids were isolated, and for each clone 1 µl was used for sequencing forward and reverse strands as described in Blattner [14] using primers pJET-F and pJET-R. For eight diploid individuals six to ten clones were sequenced to test if TOPO6 is a single-copy locus, while for all polyploids an average of 12 clones each were sequenced.

Data analyses

Manual editing of sequences and multiple sequence alignments were performed with Geneious Pro v5.4 [20] followed by manual adjustments of the alignments. Most sequences obtained by cloning of amplicons of regular Taq polymerase from single individuals had very similar sequences, differing in one to ten mutations (on average two), which were not shared by more than one clone. These differences are very likely PCR errors from the Taq polymerase that occur during cloning or sequencing. In these cases consensus sequences for highly similar sequences were created in order to reduce the number of singletons in the alignment.

Chimerical sequences can be the result of natural recombination between alleles of orthologous or homoeologous genes and/or PCR-mediated recombination. PCR recombination occurs when partially extended PCR fragments function as a primer to amplify divergent sequences [21]. Allopolyploid species are especially prone to the formation of chimerical sequences due to the presence of two or more homoeologous copies. Bifurcating phylogenetic trees cannot represent precisely the evolutionary histories of recombinant sequences and the presence of chimerical sequences disturbs analysis algorithms, as they combine signals from different phylogenetic groups. Automated methods included in Rdp3 [22] to account for recombination events were used but the results were not conclusive due to the high number of PCR-mediated mutations present in the raw data set. Therefore, sequences were thoroughly inspected by eye to identify sequences showing combinations of polymorphic sites present in different alleles (Fig. S1). The recombinant sequences were excluded from the data set prior to the analysis. In cases of identical sequences derived from the same individual only one sequence was included in the data analysis.

After a preliminary analysis, out of the 341 individuals sequenced, a reduced dataset representative of all the diversity found was used for the different analyses. Thus alleles shared by more than one individual per species were included only once. The final alignment consisted of 278 sequences. This dataset (File S1) and a subset containing only the sequences derived from diploid species and cytotypes consisting of 109 sequences (File S2) were analyzed using parsimony and Bayesian methods. In all analyses Psathyrostachys juncea was defined as outgroup [23]. A maximum parsimony (MP) analysis was conducted in Paup* 4b10 [24] using the two-step procedure described in Blattner [14]. In an initial heuristic search with 1000 random addition sequences and TBR branch swapping the number of trees retained was restricted to five per random addition. The best trees from this search were used as starting trees in a second heuristic search utilizing TBR branch swapping, restricting the number of saved trees to 50,000. To test the statistical support of clades a bootstrap analysis with 50,000 re-samples and the fast-and-stepwise algorithm was conducted in Paup*.

For Bayesian inference (BI), different models of sequence evolution were investigated with MrModeltest version 2.3 [25]. As partitioning of the data to account for intron/exon differences did not change the outcome of an initial BI analysis we inferred an overall model of sequences evolution for the entire marker region. Among the 24 models tested, the best-fit model selected by the hierarchical Likelihood Ratio Test (hLRT; [26]) and Akaike Information Criterion (AIC; [27], [28]) was HKY+Γ, a transition/transversion model [29] with rates variation according to a gamma distribution [30]. The analysis conducted with MrBayes version 3.1.2 [31] consisted of two parallel Metropolis coupled Monte Carlo Markov chain analyses with six chains per run for 5 million generations and sampling trees every 1000 generations. The temperature parameter was set at 0.05 to obtain a value of state swap frequency within the range of 10% to 70%. The convergence of the parameters was evaluated with the standard deviation of split frequencies (<0.01) and with the program Tracer version 1.5 [32]. The topology convergence was checked using the compare function of the online application Awty [33], which plots posterior probabilities of clade support values for both runs against each other. The first 25% trees were discarded as burn-in and a consensus tree was computed in MrBayes 3.1.2.

To visualize species relationships, the final analysis was summarized in a schematic tree as used by Blattner [14]. In this scheme, the phylogeny of sequences derived from diploid species was used as a backbone, and the polyploid species were connected to the diploids according to the placement of their sequences in the complete analysis.

Results

The TOPO6 sequences obtained in this study varied in lengths between 868 and 1057 bp and were stored in the EMBL nucleotide database under accession numbers HE655746-HE656023. The alignment of 278 TOPO6 sequences was 1275 bp long and contained 367 variable sites (281 for diploids only), of which 248 were parsimony-informative (205 for diploids only). All analysis algorithms resulted in very similar tree topologies, thus only the BI trees are presented (Fig. 1 for diploids only, Fig. 2 for the complete dataset), while results of the MP analyses are available as Figures S2 and S3. Both analyses are summarized in a scheme of the TOPO6-based species and cytotype relationships within Hordeum (Fig. 3). All analyses revealed the sequences derived from Hordeum species to be monophyletic with one exception (Fig. 1): two clone sequences from a single diploid H. brevisubulatum individual (PI229753) clustered outside the Hordeum clade, together with Eremopyrum triticeum. Sequences from the four genome groups in Hordeum (H, I, Xa, Xu) were mainly found monophyletic (Fig. 1) with few exceptions: (i) the H. brevisubulatum sequences already mentioned, (ii) a tetraploid individual of H. brevisubulatum (BG156/07) having one sequence falling outside of the I clade in a polytomy together with the Xa+I clade, and (iii) two diploid individuals of H. murinum (Xu genome, [12]) with two clones clustering with H-genome H. bulbosum (derived from H. murinum PI218078) and one clone clustering with I-genome H. pubiflorum (from H. murinum BCC2002). The Hordeum clade received strong support with a posterior probability (pp) of 0.99 in BI. The H-genome sequences formed the sistergroup to the remaining species, with Xu-genome sequences grouping as sister to the clade including sequences derived from Xa- and I-genome taxa.

Figure 1

Phylogenetic tree derived from TOPO6 sequences of the diploid Hordeum taxa and eight outgroup species calculated with Bayesian inference.

Posterior probability values of the clades are indicated along the branches. Bold letters depict genome denominations following Blattner (2009). After the species name and individual number the different copies found per individual are indicated (A–C) in case of cloned sequences.

Figure 2

Phylogenetic tree derived from cloned TOPO6 sequences from diploid and polyploid Hordeum taxa and eight outgroup species calculated with Bayesian inference.

Posterior probability values of the clades are indicated along the branches. Clades containing diploid and polyploid-derived sequences are indicated to the right. Genome denominations are given in bold type.

Figure 3

Scheme summarizing phylogenetic relationships of species and cytotypes in the genus Hordeum based on TOPO6.

Diploid taxa were drawn directly at the tree, while tetra- and hexaploids were connected by lines to their inferred parental taxa. Dashed lines indicate uncertainties and double lines depict an autopolyploid origin. Colors refer to the ploidy level of the taxa (tetraploid blue lines, hexaploid green lines).

In the H-genome clade (section Hordeum), sequences of both cytotypes (2× and 4×) of H. bulbosum clustered together (Fig. 2) in one strongly supported clade (1.0 pp). Between one and three similar sequences were recovered per individual (Figs. 1 and 2), indicating either duplication of the TOPO6 locus in this species or strong allelic diversity within individuals. We did, however, retrieve no sequences indicating that a taxon outside the extant diploid of H. bulbosum contributed to the formation of the tetraploid, which confirms the autopolyploid origin of this cytotype.

The Xu-genome group (section Trichostachys) of the H. murinum taxon complex consisted of one clade (1.0 pp) in the analysis of diploids (H. murinum subsp. glaucum) and two clades when polyploids were included (Figs. 1 and 2). This second clade was formed only by sequences derived from polyploids, indicating the existence of homoeologues not occurring in extant diploid individuals [12].

In the Xa-genome clade (section Marina), sequences clustered in three strongly supported groups (Fig. 2): (i) H. marinum sequences (1.0 pp), (ii) H. gussoneanum (2× and 4×, type B), H. capense (type A), H. secalinum (type A) and hexaploid H. brachyantherum (type C) with 0.98 pp, and (iii) H. gussoneanum sequences derived from tetraploids only (type A, 1.0 pp). As before, this latter group indicates homoeologues that do not occur in extant diploid individuals of section Marina.

The I-genome group (section Stenostachys) consisted of the Asian species H. brevisubulatum and H. roshevitzii being successive sister taxa to the large clade of sequences from Asian H. bogdanii together with all sequences derived from American species (Fig. 1). In this latter clade sequences of single species and species groups were arranged along a large polytomy. Hordeum bogdanii and H. californicum sequences as well as sequences of the three closely related Patagonian species H. comosum, H. patagonicum, and H. pubiflorum occured directly along this polytomy or formed clades of sequences derived from single species, while sequences from the other species formed mixed clades. Thus, H. chilense grouped together with H. flexuosum, H. cordobense with H. muticum, and H. euclaston with H. intercedens, H. pusillum, H. erectifolium, and H. stenostachys (Fig. 1). The sequences derived from H. bogdanii possessed a 33 bp region (alignment positions 925–957) shared by all Old World Hordeum species and absent in the American species (Fig. S4).

In the tree including polyploid-derived sequences (Fig. 2) in the I-genome group a clade (1.0 pp) consisting only of sequences from the three cytotypes (2×, 4× and 6×) of H. brevisubulatum was sister to a large clade with a basal trichotomy. Its first clade (1.0 pp) contained only sequences of H. capense and H. secalinum (type B) without any sequences from an extant diploid species. The second clade (0.99 pp) consisted of sequences of tetraploid and hexaploid H. brevisubulatum together with sequences originating from the diploid species H. roshevitzii, the tetraploid species H. jubatum, H. brachyantherum, H. guatemalense, H. tetraploidum, and H. fuegianum, and the hexaploid species H. arizonicum, H. brachyantherum, H. lechleri, H. parodii, and H. procerum. In the third clade, in addition to the diploid-derived sequences, Asian polyploid H. brevisubulatum and all American polyploids grouped mostly together with specific diploids into smaller subclades. Thus, H. tetraploidum (4×), H. lechleri (6×) and H. parodii (6×) fell in a clade with diploid H. pubiflorum and, augmented by H. fuegianum (4×), also in the group of sequences along the basal polytomy. Hordeum parodii (6×) grouped together with diploid H. flexuosum, H. procerum (6×) with diploid H. cordobense, H. arizonicum (6×) with diploid H. pusillum, H. depressum (4×) with diploid H. intercedens, and H. depressum (4×) and H. brachyantherum (6×) also with diploid H. californicum. Finally, there is a clade consisting of sequences found in H. brachyantherum (4×), H. guatemalense (4×), H. jubatum (4×) and all American hexaploid species. In this latter clade no sequences derived from any extant diploid species could be found.

For some tetraploid Iranian individuals of H. brevisubulatum we got no indication that other taxa than diploid H. brevisubulatum contributed TOPO6 sequences, which indicates an autopolyploid origin of these individuals.

The TOPO6 alignment revealed the presence of two insertions of transposable elements (TE). Thus, a TE of variable size ranging from 78 bp (in Secale) to 123 bp (in Triticum) is present at alignment positions 968–1092 of the outgroup species. BLAST searches of this element retrieved sequences featuring a Stowaway miniature inverted repeat transposable element (MITE). All Hordeum species plus Dasypyrum villosum and Eremopyrum triticeum were missing this element. In addition, the H. bulbosum accession BCC2018 had an allele with an insertion of 162 bp similar to a Stowaway MITE annotated as Thalos (EMBL sequence accession number AF521177.1, position 76813–76977; 92% similarity) that is located in a TA target site (alignment positions 709–872).

Recombinant clones of TOPO6 and comparison of two polymerases

Among the 945 clone-derived sequences obtained for this study, 365 from 19 tetra- or hexaploid individuals could be analyzed for the influence of proof-reading versus standard DNA polymerases. Separate PCRs were conducted for these individuals, with both standard and proof-reading enzymes. An example of the resulting sequences is shown in Figure S1. For this subset, Taq resulted in an average of 40% (SD = 0.24) of chimerical sequences, while the Phusion Hot Start DNA Polymerase resulted in only half of this proportion (22%, SD = 0.19). This difference was significant (paired Student's t-Test, p = 0.01) when testing for a higher proportion of chimerical sequences with the regular Taq than with the Phusion enzyme. Concerning only the tetraploids (10 individuals), we found a significant difference (paired Wilcoxon signed rank test, p = 0.04) with the same alternative hypotheses. In the case of hexaploids (nine individuals), no significant difference was found (paired Student's t-Test). This result might be due to the small amount of comparisons and/or to the presence of more true chimerical sequences present in the genome of hexaploid individuals.

Discussion

Our phylogenetic analysis revealed TOPO6 to be a single-copy locus in most diploid Hordeum species, as we found regularly only two different sequences per diploid individual, which we interpret as allelic variation. In a few cases we detected more than the expected maximum of two, four and six alleles in diploid, tetraploid and hexaploid individuals, respectively. However, except for three surprising cases (H. murinum H00219 and PI218078 and H. bulbosum BCC2018), these were very similar to each other so that we cannot safely discern if they stem from gene duplication or are artifacts originating during DNA amplification and sequencing. However, as using a proof-reading DNA polymerase in PCR greatly reduced the number of such sequence polymorphisms within individuals (Fig. S1) we assume that these are essentially PCR artifacts.

The phylogenetic relationships obtained from TOPO6 are mainly congruent with previous work (reviewed in Blattner [9]) although they deviate from the total evidence phylogenetic tree of Hordeum [9]. This is not an unexpected result, as up to now no single marker region was able to ‘correctly’ resolve all species relationships among diploid Hordeum species [9], [34]. The major difference (Fig. 1) is the non-monophyly of subgenus Hordeum due to H. murinum being sister to subgenus Hordeastrum instead of the H-genome taxa, yet with weak statistical support. Minor differences, although no incompatibilities, concern the positions of (i) H. bogdanii and (ii) H. californicum, which group in the large polytomy of the New World species instead of H. bogdanii being sister to H. californicum plus the clade of the mainly South American taxa, and (iii) the non-monophyly of sequences derived from the three closely related Patagonian species H. comosum, H. patagonicum, and H. pubiflorum. Sequences of these species also group along the large polytomy in Figure 1, providing no hard contradiction to monophyly of this group. As the TOPO6 sequences derived from polyploids mostly cluster with sequences obtained from specific diploid species in phylogenetic analyses, they enable the identification of the parental species involved in polyploid formation, which is the major goal of this study.

Inference of extinct diploid progenitors of allopolyploids

Four statistically well-supported and genetically distinct clades in the phylogenetic trees contain only sequences derived from polyploid species but lack sequences of a currently existing diploid taxon: (i) in H. murinum (1.0 pp), (ii) in H. gussoneanum (1.0 pp), (iii) for H. capense/H. secalinum in the I-genome group (1.0 pp), and (iv) within the H. californicum group (1.0 pp). As we included all taxa of Hordeum in this study, mostly with multiple individuals representing the distribution areas of the species, we can safely interpret these sequences as the footprints of diploid species, which contributed their genome to allopolyploid taxa in the past and went extinct sometime after polyploid formation [35]. These data support similar findings from ITS and EF-G sequences for H. murinum [12], [14], [36] and ITS sequences for H. californicum-related taxa [14], EF-G data for tetraploid H. gussoneanum [37], and are compatible with the results of analyses of the HTL gene [38] and ITS [14] in H. gussoneanum. For H. capense/H. secalinum the identification of an extinct diploid progenitor in addition to H. gussoneanum is new. Based on the position of the TOPO6 sequences between the clades formed by diploid H. brevisubulatum and H. roshevitzii (Fig. 2) we conclude that this extinct taxon belonged to the Central Asian group of Hordeum species, i.e. series Sibirica. Extinct progenitors of allopolyploids have been previously inferred [14], [35], [39]–[41]. However, the extent of this phenomenon is currently unclear, as the inclusion of single or few individuals as representatives of species is still common in phylogenetic studies (e.g., [13], [34]). This does not allow for the discerning of closely related species from conspecific individuals with high intraspecific genetic variation. Thus, Petersen and Seberg [10] obtained one type of DMC1 sequences found in H. capense and H. secalinum as sistergroup of H. brevisubulatum while sequences of the second type grouped in a polytomy together with the sequence of H. gussoneanum. Accordingly, they interpreted this topology as an indication for H. gussoneanum and H. brevisubulatum being the parents of the tetraploids. The strong differentiation between their H. capense/H. secalinum and H. brevisubulatum sequences in DMC1 are, however, completely compatible with the scenario we propose here, i.e. that the genepools of diploid H. brevisubulatum and the extinct progenitor of H. capense/H. secalinum belonged to separate taxa in the past. Particularly as neither DMC1 nor TOPO6 showed a comparably large differentiation for the tetraploid's sequences derived from H. gussoneanum, which should be the case if differentiation of the homoeologues evolved only after polyploid formation.

Due to insufficient state of data on such extinct species it is currently not possible to infer if extinction rates for diploids generally rise after they contribute to polyploid formation and, thus, increase competition in their habitats [42] or if we see the normal rate of background extinction in Hordeum. The ratio of 3∶1 of Old versus New World extinct species maintained in polyploids fits with the proposed generally higher Pleistocene extinction rates in Eurasia in comparison to the Americas inferred from the distribution of missing chloroplast haplotypes in Hordeum clades [43].

Polyploid species of the Old World

In H. bulbosum, occurring in the Mediterranean and adjacent Southwest Asia, two cytotypes (2×, 4×) exist. All sequences but one (BCC2018_A) derived from this species are quite similar and group in a single strongly supported clade (1.0 pp), while the single outlier is sister to this group. For the diploid individuals we found, as expected, up to two TOPO6 alleles and also in the tetraploid no more than two alleles were detected. Finding only one kind of sequence in a tetraploid could result from the loss of one copy from the genome, gene conversion, PCR-drift, limited clone sampling or autopolyploidy. As we included multiple individuals of the tetraploid in our study we do not expect that all would behave in the same way regarding technical shortcomings. Thus, our result supports the long-standing assumption that the tetraploid cytotype of H. bulbosum is of autopolyploid origin [44]. The peculiar position of the second type (A) of individual BCC2018 could indicate ancient introgression and/or incomplete lineage sorting [43] or pseudogenization of one TOPO6 copy.

In the H. murinum taxon group two clades of TOPO6 sequences were obtained. As already inferred by Jakob and Blattner [12] and Tanno et al. [36] in detailed studies of these taxa and cytotypes all polyploids are of allopolyploid origin involving extant and extinct species from within the Xu-genome group.

As discussed before, in the Xa-genome clade sequences obtained from H. marinum and H. gussoneanum (2×, 4×) formed three distinct clusters. The two copies of the tetraploid cytotype appeared in two of these clusters, one being exclusive to this cytotype (type A), which we interpret as an indication for an extinct diploid progenitor. Although H. gussoneanum was also involved in the evolution of three allopolyploid taxa combining I and Xa genomes, only extant H. gussoneanum contributed in these cases. One group of TOPO6 sequences of the two tetraploid sister species H. capense and H. secalinum was clustering with the sequences derived from diploid H. gussoneanum, pinpointing this taxon as one parental species and are thus in accord with previous analyses [10], [14], [45], [46]. We found, however, no indications for a contribution of H. marinum to these tetraploids, as proposed by Taketa et al. [47] based on cytogenetic and Jakob and Blattner [43] on chloroplast data. Our data also confirm that H. capense and H. secalinum are very closely related, contrarily to Baum and Johnson [48], but clearly separated taxa, which most probably speciated after long-distance dispersal of H. secalinum from Europe to South Africa resulting in the geographically isolated H. capense [46].

Diploid H. gussoneanum introduced into North America contributed to the formation of the hexaploid cytotype of H. brachyantherum (haploid genome composition IIXa) via hybridization with H. brachyantherum, most probably in historic times. This result confirms previous observations [11], [14], [43], [45]. Our TOPO6 sequences indicate that H. gussoneanum individuals originating from Spain (BCC2005) might have formed the Californian population (JB157) and contributed to H. brachyantherum (6×), occurring only in the Californian Bay area.

Hordeum roshevitzii, a key species in the evolution of tetraploid Hordeum species

In addition to polyploid Central Asian H. brevisubulatum (discussed in detail below) all American tetraploid species with the exception of H. depressum, viz. H. brachyantherum, H. fuegianum, H. guatemalense, H. jubatum, and H. tetraploidum, carry one TOPO6 type grouping with the sequences of H. roshevitzii, a diploid species endemic to Central Asia. The second sequence type derived from these species clustered with different diploid species of the American clade. North American H. brachyantherum and H. jubatum and Central American H. guatemalense thus have the roshevitzii-like TOPO6 copy plus one of the extinct species identified within the H. californicum clade. In contrast to results from an ITS analysis [14] we found no indication for introgression of H. intercedens in H. guatemalense. In South American H. tetraploidum and H. fuegianum, in addition to the roshevitzii-like type, TOPO6 copies close to that of the three Patagonian species H. comosum/H. patagonicum/H. pubiflorum occur. For H. tetraploidum, the sequences seem to hint towards a polyphyletic origin, as a group of individuals possess a different TOPO6 copy derived from H. pubiflorum. From our data we cannot discern if H. fuegianum evolved from H. tetraploidum through a speciation event on the tetraploid level or if both taxa evolved through independent allopolyploidization involving the same parental species. To resolve this we would need a higher number of informative characters in DNA sequences and much more individuals of both species included in an analysis.

Overall, H. roshevitzii appears as the key species in the evolution of the tetraploid species of the New World. Our data confirm findings of Blattner [14] using ITS, where H. roshevitzii formed a clade with American allopolyploid species, and also cytogenetic data (FISH) that infer a contribution of H. roshevitzii to American allopolyploids [49]. Surprisingly, Wang and Sun [13], based on DMC1 sequences, did not find a contribution of H. roshevitzii to American polyploid Hordeum taxa. Instead they detected in some American tetraploid and hexaploid species (H. jubatum, H. fuegianum, H. tetraploidum, and H. arizonicum) DMC1 copies with close relationships to other Triticeae genera (Taeniatherum and Pseudoroegneria). The use of single individuals as representatives for taxa in this study makes it impossible to infer reasons for these differences, i.e. if this is a general feature of the polyploids or a peculiarity of the DMC1 locus for single individuals.

In a geographic context the occurrence of roshevitzii-like sequences in North American polyploids indicates a second colonization event from Asia to North America after the initial establishment of bogdanii-like Hordeum diploids on that continent [46]. The roshevitzii-like sequences of the two South American tetraploids H. tetraploidum and H. fuegianum require either initial colonization of South America by H. roshevitzii and its later extinction after the formation of the tetraploids or the introduction of this sequence types through a polyploid (H. jubatum, see below) and introgression of the alleles via hybridization.

The extinct californicum-like taxon was a key species in the evolution of American polyploids

The extinct close relative of H. californicum, to which we refer informally as H. humatio-californicum (due to the fact that it is closely related to H. californicum but that its genome today can only be found ‘buried’ in polyploid taxa), contributed its genome to tetraploid H. brachyantherum and H. jubatum and all New World hexaploids, i.e. H. arizonicum from North America and South American H. lechleri, H. parodii and H. procerum. Therefore, its importance for the evolution of Hordeum polyploids is comparable to that of H. roshevitzii. The American hexaploids all show essentially the TOPO6 homoeolog pattern of tetraploid H. jubatum. In addition each hexaploid has a third copy clustering with American diploid species. Thus, the formation of these hexaploids can be explained by hybridization of H. jubatum with (i) diploid H. pusillum resulting in H. arizonicum, with (ii) H. cordobense resulting in H. procerum, with (iii) H. pubiflorum resulting in H. lechleri, and with (iv) H. flexuosum or, geographically less likely, H. chilense resulting in H. parodii. While the first two combinations seem plausible regarding the partly overlapping distribution areas of the species involved, and the third also for the overall high morphological similarity of H. lechleri with its proposed progenitors, in H. parodii we found only one individual with the H. humatio-californicum sequence, whereas four individuals possess a TOPO6 copy related to H. comosum/H. patagonicum/H. pubiflorum. In this latter case, it might also be that H. tetraploidum instead of H. jubatum contributed to hexaploid formation and later-on introgression with H. jubatum or one of the other hexaploids contributed to additional genomic diversity. Alternatively, also a polyphyletic origin of H. parodii is compatible with the data. Chloroplast sequences do not contribute to the clarification of this topic, as apart from H. lechleri, all South American hexaploids possess chloroplast haplotypes derived from their South American diploid progenitors, i.e. H. parodii chloroplast haplotypes are shared with diploid H. patagonicum and H. pubiflorum and tetraploid H. tetraploidum [43].

In any case we have to assume that H. jubatum was present in South America for a long enough time to allow the evolution of at least three hexaploid taxa and their expansion to their extant partly very large and mainly allopatric distribution areas. This is in contradiction to the present assumption of a natural distribution of H. jubatum only in northwestern North America and northeastern Siberia [8] and its introduction into other areas of the world as ornamental in historic times. From the data obtained from the South American polyploids we infer the natural occurrence of H. jubatum in southern South America prior to European settlements and trade routes in this area. Thus, the scattered stands of this taxon in the grasslands of Central Argentina might well have originated by ancient bird-mediated [46] long-distance dispersal of this taxon from North America [43].

Hordeum depressum, an All-American tetraploid

Hordeum depressum possesses a TOPO6 copy (A) derived from H. californicum and a copy (B) clustering with diploid H. intercedens and H. euclaston. As it obtained no roshevitzii-like TOPO6 copy it is the only ‘purely’ American tetraploid species. Hordeum intercedens from southwestern California and adjacent Mexico phylogenetically groups within South American H. euclaston [50]. Pronounced ecological differentiation [12] together with the geographically caused de facto reproductive isolation [46], [50] warrants the recognition as two independently evolving lineages, i.e. separate species. In contrast to Wang and Sun [13] who assumed that H. euclaston is a parent of H. depressum, we propose that in the frame of the geographical co-occurrence of H. californicum and H. intercedens in southern California these species are much more likely to hybridize and thus to have contributed to the evolution of the tetraploid.

Hordeum brevisubulatum, a complex group

The taxon complex of H. brevisubulatum comprises five described subspecies from a large geographic area reaching from western Turkey to northeastern China with diploid, tetraploid and hexaploid cytotypes. Identification of the different subspecies is delicate especially for herbarium samples and materials from areas where subspecies overlap, thus we decided to consider only the ploidy level and the country of origin. The species is assumed to be of autopolyploid origin [14]. We were able to discern two major groups, with Iranian individuals found to be molecularly different from individuals from the rest of the species' distribution area. With one exception (H00312) the tetraploid individuals of H. brevisubulatum originating from Iran appeared to be autopolyploid, as the TOPO6 sequences derived from these individuals clustered with the diploids only. Surprisingly, polyploid individuals originating from Siberia did not have a diploid brevisubulatum-like copy but instead one type of sequences derived from these eastern individuals clustered with H. roshevitzii. The other copy was either found in a peculiar position on a polytomy with the clades formed by Xa and I-genome sequences (BG156_07) or in the New World clade. In this latter clade they cluster together with sequences recovered from a tetraploid accession from Kirgizstan and a Tajik individual. For the hexaploid cytotypes, the individuals from Iran had a diploid brevisubulatum-like copy and the other one was clustering in the New World clade together with other H. brevisubulatum sequences. The hexaploid Tajik accession (BCC2016) had one sequence falling in the H. roshevitzii clade and one in the tetraploid/hexaploid H. brevisubulatum clade embedded in the New World clade. The H. brevisubulatum sequences present in this latter clade are similar to H. bogdanii, especially regarding the absence of the 33 bp long deletion characteristic for the American taxa.

To summarize our findings regarding H. brevisubulatum, according to the TOPO6 phylogeny, only the Iranian H. brevisubulatum tetraploids are of autopolyploid origin, while in the remaining distribution area allopolyploids occur, which do not even include the diploid's TOPO6 type. This species complex thus seems polyphyletic and/or exhibits signs of long-term interspecific hybridization with the three diploid Asian taxa H. brevisubulatum, H. bogdanii and H. roshevitzii. Moreover, H. brevisubulatum is the only species having one diploid individual with a TOPO6 sequence clustering with the outgroup Eremopyrum triticeum, indicating introgression from outside of Hordeum [13], [51]. What we cannot estimate is if the genetic diversity found here represents the natural state of the species' lineages or if hybridization occurred during reproductive cycles in germplasm repositories. Obligate outbreeding of H. brevisubulatum [8] might facilitate introgression in comparison to most other Hordeum species.

PCR-recombination results in chimerical sequences

Conducting this study, we tested for the influence of two different DNA polymerases on the proportion of chimerical sequences, combining parts of different TOPO6 types occurring in Hordeum. Lahr and Katz [52] found that for genes of the major histocompatibility complex proof-reading DNA polymerases greatly reduce the proportion of recombinant sequences originating during PCR. In Hordeum we found the same effect, however, it seemed to vary according to the ploidy level, i.e. in tetraploid individuals the reduction of chimerical amplicons was much higher than in hexaploids. There are two possible explanations for this result. The first reason may be due to a lack of statistical power, as fewer comparisons were conducted (we did not increase the amount of screened colonies for hexa- in comparison to tetraploids). The second might be due to biological reasons, as recombinant copies might already be present in the genome. Cronn et al. [53] also found recombinant sequences of low-copy number genes in allopolyploid cotton species, which indicates that this phenomenon is not restricted to Hordeum. Although low-copy nuclear genes are very promising to reconstruct species phylogenies especially in polyploids [54], [55], the problem of recombinant sequences limits the usage of such markers. The origin of chimerical sequences during PCR amplification seems inherent to the type of sequences and to the use of universal primers, as they uniformly amplify all homoeologues, resulting in a mixture of amplicons. To minimize this problem one can design homoeolog-specific primers [10], [41], [53]. This means, however, that sequence information for the homoeologues has to be present in advance and prevents naïve exploring of allele diversity. Another solution, using single-molecule (sm) PCR [56], was employed by Marcussen et al. [57] to disentangle reticulate evolution in Viola. The smPCR seems particularly suitable for high-polyploid species and reduced sampling size. Nevertheless, from our experience in Hordeum we suggest that the use of a proof-reading DNA polymerase, probably together with low initial DNA concentration [52] in PCR, can reduce potential ambiguities regarding artificial recombinant amplicons. Although, we are well aware that this cannot supersede a careful inspection of the data, particularly when higher ploidy levels are involved.

Conclusions

Using cloned sequences of TOPO6, a low-copy nuclear region, in a comprehensive framework including all Hordeum species mostly with several individuals, covering the geographic distribution of the species, we were able to infer parental relationships of Hordeum polyploids. The phylogenetic hypothesis presented here (Fig. 3) brought several new insights and supported other earlier data. Thus, it is likely that a close relative of Asian H. bogdanii was the starting point for the evolution of American diploid species. Diploid H. roshevitzii together with an extinct close relative of H. californicum and tetraploid H. jubatum were pivotal species for the evolution of the Hordeum allopolyploids. The involvement of H. jubatum in the formation of South American polyploids necessitates the presence of this taxon in South America well before the onset of European settlement. Thus we propose that also Central Argentina belongs to the natural distribution area of this species, resulting in a disjunct Northern–Southern Hemisphere distribution of H. jubatum. We were able to analyze the status of two putative autopolyploids, confirming autopolyploidy for H. bulbosum and tetraploid Iranian populations of H. brevisubulatum, while the latter taxon shows otherwise very complex and still poorly understood allopolyploid patterns. This species complex as well as some polyploids in South America might result from multiple independent origins or had a long history of hybridization and introgression. The use of proof-reading DNA polymerase in PCR can reduce phylogenetic noise when analyzing polyploid sequences by cloning. More in-depth analyses have to be performed to resolve the still unclear parental relationship, for example in some cases using more loci to arrive at a higher resolution for closely related species groups (in South America) or by including much more geographically representative individuals from natural populations for species complexes (particularly in H. brevisubulatum).

Examples of chimeric sequences together with non-recombinant sequences derived from a tetraploid individual of (JB048C2B). The striped alignment shows only polymorphic sites of all 16 clone-derived sequences (JB048C2Bb-q) and the two consensus sequences used in the analysis (H. tetraploidum JB048C2B_A and B). Sequences b–h were obtained using standard DNA polymerase and sequences i–q using proof-reading DNA polymerase in PCR. Sequences c, e, f and h show mosaic patterns indicated below the sequences as “A type” or “B type”, referring to the sequences used in the analysis. For this individual no chimerical sequences were retrieved with the proof-reading DNA polymerase against four obtained via regular DNA polymerase. Also the amount of singleton SNPs (PCR errors) found in the sequences differs between both polymerases.

(PDF)

Click here for additional data file.

Strict consensus tree of 50,000 most parsimonious trees (L = 374 steps, CI = 0.84, RI = 0.96) from an analysis of sequences derived from diploid taxa and eight outgroup species. Numbers along branches depict bootstrap values (%) of major clades of the tree derived from a ’fast-and-stepwise’ analysis of 50,000 bootstrap re-samples. Psathyrostachys juncea was defined as outgroup taxon in the analysis.

(PDF)

Click here for additional data file.

Strict consensus tree of 50,000 most parsimonious trees (L = 536 steps, CI = 0.80, RI = 0.97) from an analysis of sequences derived from di- and polyploid taxa and eight outgroup species. Numbers along branches depict bootstrap values (%) of major clades of the tree derived from a ‘fast-and-stepwise’ analysis of 50,000 bootstrap re-samples. Psathyrostachys juncea was defined as outgroup taxon in the analysis.

(PDF)

Click here for additional data file.

Part of the alignment showing a 33 bp deletion (alignment positions 925–957) synapomorphic for sequences derived from diploid New World species. Sequence deletion occurred at a five basepair direct repeat (TACAC) flanking the deleted region (arrows). The absence of the deletion in H. bogdanii together with its presence in all American diploid species indicates that not H. bogdanii itself but a close relative of this species was the initial starting point for the colonization of the Americas by Asian Hordeum.

(PDF)

Click here for additional data file.

Detailed information for all individuals analyzed in this study.

(PDF)

Click here for additional data file.

Alignment of di- and polyploid-derived sequences.

(FASTA)

Click here for additional data file.

Alignment of diploid-derived sequences.

(FASTA)

Click here for additional data file.