DNA Barcodes for B. distachyon, B. stacei and B. hybridum

Under the premise that a successful barcode locus should enable the recovery of monophyletic clusters corresponding to individual species [34], we found that any one of the three assayed loci (trnLF, ITS, GI) could unambiguously differentiate the two monophyletic diploid species from direct sequencing of PCR amplicons. However the identity of the allotetraploid requires combined analysis of direct trnLF and direct or cloned ITS sequences or through analysis of cloned GI sequences.

Our results demonstrate that the widely employed barcoding regions trnLF and ITS [20], [23] clearly discriminate between B. distachyon and B. stacei. Both regions showed: i) high inter- vs intraspecific distance divergences, ii) significant barcoding gaps (Table 2), iii) extremely distant monophyletic clusters in the parsimony networks (Figs. 2a, b); and iv) highly supported divergent monophyletic clades in both the NJ (Results not shown) and the Bayesian trees (Figs. 3a, b). They also comply with the requirements of feasibility and rapid and easy production of the sequences to be considered optimal barcoding molecules [20]. However, the allopolyploid nature of B. hybridum, together with its estimated recent origin (c. 1 Ma; [11]), prevents their direct use as single standard barcodes for this taxon and its two parental taxa. Our study has shown that the maternally-inherited B. hybridum trnLF haplotype sequences could have been acquired from either of the two parents (Table 1; Figs. 2a, 3a) and that the biparentally-inherited B. hybridum ITS copies (B. distachyon-like and B. stacei-like) could either have remained intact in the hybrid genome or could have converged into one or the other parental copy (Table 1, Figs. 2b, 3b). This creates the possibility of misleading results if the B. hybridum trnLF and ITS sequences had been respectively inherited from and (co-inherited but) converted into the same progenitor sequences, causing confusion between the parent and the allotetraploid taxa (e. g. Bhyb26, Bhyb30and Bhyb35 with B. distachyon, and Bhyb28, Bhyb40, Bhyb41, Bhyb 47, Bhyb49 and Bhyb105 with B. stacei; Table 1). Cloning of the ITS sequences can help to solve the uncertainty if both parental copies are detected, as demonstrated in several studied cases (e. g. Bhyb9, Bhyb10, Bhyb14, Bhyb15, Bhyb18, Bhyb19, Bhyb22, Bhyb23 and Bhyb38; Table 1). The use of the combined trnLF+ITS barcode shows high percentages of successful species discrimination among the species in the reticulate triangle using either direct trnLF and ITS sequences (93.3%) or direct trnLF and cloned ITS sequences (94% (Tables 1, S2). The barcoding would remain untractable, however, if the concerted-evolution mechanism that operates in the multicopy nuclear ribosomal genes [26], [35] had converted all the co-inherited copies into the same parental copy.

Because of the drawbacks posed by the use of these classical barcodes, we searched for an alternative nuclear locus that could unambiguously differentiate the three species. This could only be a single-copy nuclear gene that retained both parental copies in the allotetraploid without undergoing convergent evolution towards one of them. Among the several COS proposed as appropriate candidates to differentiate closely related plant species [28], [29], [30] and to discriminate among Brachypodium taxa [11], [17] we selected a 665 bp fragment of the GIGANTEA gene, one of the key regulators of flowering promotion and phase transition [36]. This GI region has proved to be a strong candidate barcode for the B. distachyon s. l. taxa based on: i) its easy amplification, cloning and sequencing; ii) its single-copy orthologous nature: iii) the accumulation of discriminating mutations between the B. distachyon and B. stacei sequences (3.8% of mean inter- vs. intraspecific distance divergence and a significant barcode gap, Table 2); iv) the common presence of the two different co-inherited parental B. distachyon-like and B. stacei-like GI sequences in B. hybridum (Table 1); and v) rarely, the presence of inter-parental recombinant sequences that could be easily detected (Table1). The genetic differences were reflected in the GI parsimony network (Fig. 2c) and in the NJ (Results not shown) and Bayesian GI (Fig. 3c) topologies that recovered, respectively, distant clusters and well supported divergent monophyletic clades for B. distachyon and B. stacei, each of them including their respective derived B. hybridum copies. Although 5 cloned sequences were sufficient to detect both parental copies in most of the studied B. hybridum samples, a few difficult samples required the screening of up to 10–16 clones (e. g. Bhyb13, Bhyb34, Bhyb35 Bhyb50) or even a larger number, like in the case of the Bhyb69 sample (58 clones), to pick up variation from both parental species. Nonetheless, one sample (Bhyb2) showed only one parental copy after a relatively intensive clonal screening (49 clones; Tables 1, S2). This implies that a larger number of GI clones should be sequenced in order to detect co-inherited copies from both parents, providing that they are still maintained in the hybrid genome.

All the above evidence supports the choice of the GI locus as an alternative or as an additional suitable barcode for discriminating among the triangle species of the B. distachyon s. l. complex. This demands the use of cloning procedures but reduces the number of surveyed loci to just one. Moreover, the percentage of successful species discrimination increases to 98.2% (Tables 1, S2), which is above than that of the combined trnLF+ITS barcode. It further complements alternative cytogenetic identifications based on genome size or chromosome counting. The choice of the best method in a given situation would depend on considerations of facilities and costs, the acceptable error rate, and a priori information on the levels of polyploids in the sample. Very likely, other single-copy genes, such as those analysed within Brachypodium that also showed both co-inherited parental copies in the derived hybrid (e. g. CAL, DGAT, SST3; [11], [17]), could also serve as barcodes for this group of taxa. Single-copy nuclear genes are not ideal universal barcodes for plants as their priming sites cannot be easily transferred to non-related groups (e. g. [37]). The GI locus has been successfully amplified and sequenced in different representatives of Pooideae (López-Álvarez & Catalán, unpublished data) and could probably be extended to all the grass family. We propose the use of single-copy genes as a suitable barcoding alternative to circumvent the problem posed by the existence of recently evolved hybrids and polyploids within specific plant groups. In the future, the use of Next Generation Sequencing (NGS) data (e. g. [38], [39], [40]), may facilitate the barcoding of problematic plant groups which contain recently evolved hybrids and polyploids. Although the availability of NGS data is still limited both taxonomically and among laboratories, its use for this purpose is rapidly increasing. In the mean time, the use of single-copy genes is the most practicable current solution for barcoding such plant groups.

Utility of the Proposed Barcoding Method

The new DNA barcoding method proposed here has direct applications to many on-going studies of the model plant B. distachyon and its close allies [2], [4]. It has great relevance to the selection of wild germplasm for genomic (http://brachypodium.pw.usda.gov) and plant breeding programs, and for ecological and evolutionary studies of wild populations [6], [11]. For this, the correct identification of the three species is crucial but still troublesome due to uncertainty in identifications based on highly variable morphological traits and on ambiguous genome sizes, which show overlapping sizes for B. distachyon and B. stacei [11], [41]. Our study has revealed several misidentifications of B. distachyon and its close relatives B. stacei and B. hybridum in germplasm collections (e. g. USDA, ABR) and inbred lines (cf. [7], [8], [11], [42]; e. g., Bsta9, Bsta42, Bsta43, Bhyb9, Bhyb10, Bhyb19, Bhyb20, Bhyb21, Bhyb38, Bhyb39, see Table 1) that likely resulted from incorrect orcein-staining chromosome counts or misleading genome size measurements. Alternatively, the misidentifications could also result from the mixed sampling of individuals or seeds of different species from admixed populations. This problem has been manifested in the failure of ‘intraspecific’ B. distachyon crossing programs, which were in fact interspecific (Magda Opanowicz and John Doonan, pers. comm.) and in unexpected results from cell wall analyses of putative B. distachyon lines, which corresponded to B. stacei or B. hybridum lines (Richard Sibout, pers. comm.). Our barcoding method overcomes these problems, providing an efficient and automatable method to discriminate among the three species.

The validity of our proposed barcoding method depends on the large genetic divergences detected between the diploid B. distachyon and B. stacei genomes for the three analysed loci (Tables 1, 2). The high number of synapomorphic mutations separating them (23, 33, and 30, respectively, for the trnLF, ITS and GI loci; Fig. 2), facilitates the immediate classification of the genomes, even from incomplete sequences (Figs. 3a, b, c). Furthermore, the three loci provide informative indels that differentiated B. distachyon and B. stacei, like the two 6-bp gaps in the trnLF locus, the two 3- and 4-nts gaps in the ITS locus, and the one 1-nt gap in the GI locus (see Table S3). Within the ITS region, the ITS2 spacer covers the two diagnostic indels and more than half of the synapomorphic markers detected within the locus (24 out of 43; Table S3), supporting the proposal that the ITS2 subregion could be used alone (Hollingstworth 2011) to barcode case study species. The correct identification of B. hybridum would always require, however, the combined use of, at least, the trnLF+ITS barcoding sequences. Our data indicate that direct PCR sequences from the two genes could discriminate B. hybridum from its two parental species in a high percentage of the cases (88.75%; Table 1). This value increases to 90.0% when the ITS products are cloned. However, as the method might not permit full resolution, due to the potential inheritance of the same parental plastid trnLF and converted nuclear ITS sequences in the hybrid (cf. [25]), the single-copy GI locus was selected as an alternative barcode for the species in the triangle. The random screening of 5 individual GI clones gave a relatively high resolution (80%) that became higher (96.7%) when up to 10–16 (and exceptionally more, e. g. 58) clones were sequenced within our surveyed samples (Table 1).

Recently, Giraldo et al. (2012) [41] proposed a new molecular method to differentiate the three taxa based on the different allelic SSR profiles of B. distachyon and B. stacei at four nuclear microsatellite loci and their additive patterns in B. hybridum. This represents an important step forward for rapid molecular identification of the species, similar to the molecular marker-based barcoding methods proposed for taxonomically complex and highly reticulate plant (e. g. [43]) and animal (e. g. [44]) groups. However, these methods could be less stable and prone to substantial changes than the sequence-based ones as the SSR allelic variation of the barcoded species might be greater than their DNA sequences (and consequently overlap) when a wider range of samples is used [45]. The discriminating SSR markers proposed by Giraldo et al. (2012) [41] were tested across a wide representation of Spanish samples and in the type specimens of the three taxa, but they were not studied in samples from other Mediterranean regions. Thus, our barcoding approach and that of Giraldo et al. (2012) [41] could be used in a complementary way (e. g. [44]) for rapid and accurate molecular identification of the ‘Brachy-complex’ taxa [2], allowing for confident identification even when unusual allelic variation renders one or other method unreliable.

Genetics and Geographical Distributions of the Three Species and the Polyphyletic Origin of B. hybridum

Our current barcoding survey of B. distachyon, B. stacei and B. hybridum samples has encompassed the whole Mediterranean region, the native distribution area of the three species [11]. One of the main findings of the study is the detection of B. stacei populations in both the western and eastern Mediterranean regions (Table 1; Fig. 1). This rare species was until recently only known from the type locality (Spain: Balearic Islands: Formentera) [11]. However, other recent studies have indicated its presence in other localities of SE Spain [7], [41] and in the Canary Islands [41]. Our analyses have confirmed most of these findings and have also revealed its presence in other western Mediterranean localities (Mallorca (Balearic Islands), S Spain, NW Morocco; Table 1, Fig. 1) where it was mislabelled as B. distachyon in the herbaria vouchers. Most notably, we have revealed the presence of B. stacei in the SW Asian-Middle East region (Iran, Israel, Lebanon, Palestine; Table 1), from which it was unknown and also misclassified as B. distachyon. Knowledge of this broader native geographical distribution area of B. stacei will be highly valuable for the selection of new ecotypes and local lines that could be used in the generation of F2 progenies to help the assembly of the newly sequenced B. stacei ABR114 genome (http://brachypodium.pw.usda.gov; John Vogel, pers. comm.). Our study has also contributed to understanding the native distribution areas of the more widely distributed species B. distachyon and B. hybridum (Table 1; Fig. 1). Both taxa are widespread in the Mediterranean region and largely overlap [2], [8], [11]. The new barcoding data confirm their presence on both sides of the Mediterranean basin, from which regions most the germplasm lines have been generated [2], [8], [41], and also report their presence in the central Mediterranean area (Table 1). This would be also a valuable source of information for the selection of new B. hybridum ecotypes and lines for the production of F2 progenies that would complement the assembly of the newly sequenced B. hybridum ABR113 genome (http://brachypodium.pw.usda.gov; John Vogel, pers. comm.), and those of B. distachyon that could be added to the resequencing project of the model plant.

Despite their abundant distributions in the Mediterranean, the intraspecific genetic diversities of the parental B. distachyon (0.5% trnLF and ITS; 0.4%GI) and B. stacei (0.1% trnLF; 0.5% ITS; 0.3% GI) sequences were low (Table 2). This was manifested in the sharing of their respective most common trnLF, ITS and GI haplotypes by individuals from populations located far apart in the circumMediterranean region (Tables 1, S2; Fig. 2). In contrast, individuals from geographically close populations, or even intraindividual clones, showed different haplotypes. Our results agree with those of Vogel and co-workers [8] and Mur and co-workers [2], based on SSR markers, which found close genetic connections between geographically distant B. distachyon populations in Turkey and between Spain and Turkey, respectively. Selfing species are expected to show low within-population and high among-population genetic diversities [46]. However, the autogamous B. distachyon and B. stacei samples show low overall geographical structuring of genetic diversity. This might be a consequence of the long distance dispersal of their seeds (cf. [8]) and the high capability of these annuals to adapt to different environmental conditions (cf. [6]). The genetic diversity of the less abundant B. stacei could be lower than that of the more widespread B. distachyon, as deduced from the proportionally fewer trnLF and ITS haplotypes detected in the former (Table 1). Both taxa show, however, some traces of geographic isolation between the western and eastern Mediterranean regions, evidenced by the detection of regional haplotypic clades (e. g. B. distachyon: western Mediterranean, Iberian, Turkish and Middle East subclades (trnLF, Fig. 3a); B. stacei: E Iberian-Balearic and Turkish subclades (ITS, Fig. 3b). The phylogeographic study of these populations is currently in progress (López-Álvarez and coauthors, unpublished results).

Another striking finding of our study is the demonstration of the existence of different directional crosses that likely gave rise to the new allotetraploid species (Tables 1, S2; Figs. 3a, b, c). In the more restricted study of Catalán and co-workers [11], all the surveyed B. hybridum individuals showed the inheritance of a B. stacei-like plastid genome, resulting from a cross with maternal B. stacei and paternal B. distachyon parents. However, our survey with larger sample sizes shows that, although the above seems to be most common cross direction, in a few cases the B. hybridum individuals are derived from a cross between maternal B. distachyon and paternal B. stacei parents (Table 2; Fig. 3a). The fact that B. hybridum plants derived from the alternate-direction crosses occurred in different Mediterranean localities (Table 1; Fig. 1) supports the multiple and polytopic origins of the allotetraploid B. hybridum. A closer inspection of the more variable ITS and GI networks and phylogenetic trees also reveals distinct relationships of the B. hybridum sequences to different parental haplotypic groups (Table 1; Figs. 2, 3) corroborating the polyphyletic origin of the B. hybridum samples. Complementary or unique parental haplotypic clusters have been found for some Iberian (GI) and eastern Mediterranean and Balearic (ITS) B. hybridum groups (Table 1; Figs. 2, 3). Furthermore, the low mean ‘interspecific’ divergence rates shown by the B. distachyon-like and B. stacei-like sequences of B. hybridum with respect to those of the two progenitors for the three studied loci (Table 2) indicate that the two genomes of the hybrid have kept the same or similar signatures as those of the ancestral genomes, supporting the recent origin of B. hybridum in the Pleistocene (cf. [11]). Additionally, the low mean ‘intraspecific’ divergence rates of the respective B. distachyon-like and B. stacei-like sequences of B. hybridum (Table 2), which are similar to the parental ones, suggests that the original genomes have remained largely intact and that the time elapsed since the hybridizations took part was a brief one. Nonetheless, the detection of some interspecific ITS and GI recombinant sequences in B. hybridum (Table 1; Figs. 2, 3) points towards the occurrence of frequent genomic rearrangements within the hybrid nucleus. This agrees with cytogenetic CCP evidence demonstrating the existence of structural rearrangements in the B. hybridum chromosomes with respect to the B. distachyon and B. stacei ones [16].

The recurrent formation of allopolyploid plant species has been largely documented in the literature [47], [48] and references therein). Their predominance over their parental diploid progenitors has been explained as the result of their higher fitness or their higher capability to colonize new habitats and new lands [49], [50]. The wide distribution of B. hybridum, which exceeds those of B. distachyon and B. stacei in their native Mediterranean region, as the only known species of the complex to have colonized other continents [9], , could be a consequence of its more genetically diverse hybrid genome and the likely recurrent origin of new hybrid variants. This could have resulted in fit and well adapted individuals that have displaced the parental species from their habitats and/or have invaded new niches [50]. Current studies are under way to investigate the recurrent origins of B. hybridum through time (López-Alvarez & Catalán, unpublished results).

Future Perspectives of the Barcoding Method for Other Brachypodium Taxa

The almost exclusively self-fertile breeding system of the cleistogamous B. distachyon [8] and of B. stacei (L. Mur, pers. comm.) resulted in highly homozygous genomes of the two diploid parental species that contributed to the heterozygous allotetraploid B. hybridum genome [41]. In a recent assessment of genetic distances between different parent-pairs of hybrid plants, Paun and co-workers [51] concluded that parental species of allopolyploids were genetically more divergent that those of homoploid hybrids. Within Brachypodium, the differences in the inter- vs. intraspecific divergence values between the B. stacei and B. distachyon sequences were significant (Table 2). Catalán and co-workers [11] also found significant differences in the evolutionary rates of the B. stacei and B. distachyon ITS sequences, the former being significantly higher than the later. The salient features of the two distinct genomes were demonstrated through incompatible cross-GISH hybridizations [13], [14]. Their genomic divergences could have triggered the allopolyploidization process that resulted in the B. hybridum populations, and the long isolation of the two parental taxa has facilitated the detection of the proposed trnLF - ITS - GI barcoding method to distinguish the parents and the hybrid.

The usefulness of our DNA barcoding approach at the generic level could however be less successful among recently evolved taxa, like the core-perennial group of Brachypodium species, due to their close relationships [11], [52]. No significant differences in plastid trnLF and nuclear ITS sequences were detected between pairs of long rhizomatous Brachypodium species, nor between them and B. distachyon [11]. They were found, however, between the ancestral short-rhizomatous B. mexicanum and annual B. stacei taxa. Widespread geographical sampling would be required to test the utility of the trnLF and ITS barcodes within Brachypodium as a whole. Regarding GI, all the six analysed Brachypodium species [17] showed different sequences and copies, with copy numbers related to their ploidy levels. The apparently more-promising GI barcode should also be evaluated within a wide geographical and taxonomical sample of Brachypodium representatives. Brachypodium has been proposed as a model plant genus for temperate grasses [15], based on the overall small genome size of its members, their compact genomes and an extensive reticulate evolutionary and polyploid history [16]. Diverse stable species (e. g. B. phoenicoides, 2n = 4x = 28) and cytotypes (e. g. B. pinnatum 2n = 4x = 28) are of hybrid origin [16], [17] and most of the polyploids (e. g. B. mexicanum, B. retusum) are of suspected hybrid origin. Further research is currently under way to find a universal barcoding system for Brachypodium.

Sampling

A total of 210 samples (56 of B. distachyon, 43 of B. stacei and 111 of B. hybridum) were included in the study (Table 1). Those samples corresponded to inbred lines generated at CRF-INIA, INRA, USDA, and Aberystwyth (ABR), Alcalá de Henares (UAH), Jaén (UJA), Politécnica de Madrid (UPM), Tel-Aviv (TAU) and Zaragoza (Unizar) Universities and to new germplasm accessions collected over their entire native distribution areas in the Mediterranean region (Fig. 1). The identity of most of the samples was tested through DAPI-staining chromosome counting of the studied materials, which has proved to be the most accurate cytogenetic method to differentiate the three taxa. This was coupled with other identifications based on flow cytometry measurements (genome size) and anatomical stomata-leaf guard cell length measurements which separated, respectively, B. hybridum from the diploids, and all the three species [11]. We used this information as an a priori method to validate the resolution power of the proposed barcodes for the discrimination of the three species amplified with highly conserved primers.

DNA Barcode Sequences of Plastid and Nuclear Genes

Three loci were tested as potential tools for effective DNA barcoding of B. distachyon, B. stacei and B. hybridum, the maternally inherited plastid trnLF region (trnL(UAA) intron - trnL(UAA) exon - trnL(UAA)/trnF(GAA) spacer) and the biparentally inherited nuclear multicopy ribosomal internal transcribed spacer ITS region (ITS1-5.8S-ITS2) and single-copy GIGANTEA (GI) gene. Total DNA was extracted from dried leaf tissue using a modified CTAB method of Doyle and Doyle [53]. The plastid trnLF region was amplified and sequenced from direct PCR products using the universal primers ‘c’ forward (5′-CGAAATCGGTAGACGCTACG-3′) and ‘f’ reverse (5′-ATTTGAACTGGTGACACGAG-3′) of Taberlet and co-workers [54]. PCR products were purified using ExoSAP-ITTM (USB Corporation, Cleveland, OH) and sequenced in both directions by cycle-sequencing using the Big-Dye version 3 chemistry (Perkin-Elmer), with a Prism 3100 Genetic Analyzer (ABI). The nuclear multicopy ITS region was amplified using primers ITSL forward (5′-TCGTAACAAGGTTTCCGTAGGTG-3′) and ITS4 reverse (5′-TCCTCCGCTTATTGATATGC-3′) optimized for grasses [55]. A 665 bp portion of the nuclear single-copy GI gene was amplified with primers GIGIE1F (5′-TATGTCWGYNTCAAATGGGAAGTGG-3′) and HGIE5R (5′-AACTTTRAAGATTGGCCTRTTGTRGTGA-3′) designed for Brachypodium [56]. Due to the, respectively, plausible and known existence of multiple ITS and GI copies in the B. hybridum samples, and the potential existence of more than one ITS copy in the B. distachyon and B. stacei samples, all GI and most ITS amplified products were cloned and sequenced, aiming to detect their potential intraindividual copy number variation. Sixty one amplified ITS (24) and GI (37) fragments were cloned separately into a pGEM®-T Vector System I cloning vector (Promega, USA) following the manufacturers’ instructions. These were transformed into Escherichia coli JM109 competent cells. Five colonies of each individual sample were randomly picked for a first ITS and GI screen, and each clone was sequenced using M13 forward (5′-GTTTTCCCAGTCACGAC-3′) and reverse (5′-CAGGAAACAGCTATGAC-3′ ) primers following the same procedure as for direct-PCR products. In some ambiguous cases, up to 10–16 (58) clones per individual sample had to be sequenced to detect the copy number variation (see Results). The characteristics of the amplification conditions for each barcoding gene are indicated in Supplemental methods (see Methods S1).

Data Analyses