Evolution and taxonomic split of the model grass Brachypodium distachyon

Abstract

Background and Aims

Brachypodium distachyon is being widely investigated across the world as a model plant for temperate cereals. This annual plant has three cytotypes (2n =  10, 20, 30) that are still regarded as part of a single species. Here, a multidisciplinary study has been conducted on a representative sampling of the three cytotypes to investigate their evolutionary relationships and origins, and to elucidate if they represent separate species.

Methods

Statistical analyses of 15 selected phenotypic traits were conducted in individuals from 36 lines or populations. Cytogenetic analyses were performed through flow cytometry, fluorescence in situ hybridization (FISH) with genomic (GISH) and multiple DNA sequences as probes, and comparative chromosome painting (CCP). Phylogenetic analyses were based on two plastid (ndhF, trnLF) and five nuclear (ITS, ETS, CAL, DGAT, GI) genes from different Brachypodium lineages, whose divergence times and evolutionary rates were estimated.

Key Results

The phenotypic analyses detected significant differences between the three cytotypes and demonstrated stability of characters in natural populations. Genome size estimations, GISH, FISH and CCP confirmed that the 2n = 10 and 2n = 20 cytotypes represent two different diploid taxa, whereas the 2n = 30 cytotype represents the allotetraploid derived from them. Phylogenetic analysis demonstrated that the 2n = 20 and 2n = 10 cytotypes emerged from two independent lineages that were, respectively, the maternal and paternal genome donors of the 2n = 30 cytotype. The 2n = 20 lineage was older and mutated significantly faster than the 2n = 10 lineage and all the core perennial Brachypodium species.

Conclusions

The substantial phenotypic, cytogenetic and molecular differences detected among the three B. distachyon sensu lato cytotypes are indicative of major speciation processes within this complex that allow their taxonomic separation into three distinct species. We have kept the name B. distachyon for the 2n = 10 cytotype and have described two novel species as B. stacei and B. hybridum for, respectively, the 2n = 20 and 2n = 30 cytotypes.

INTRODUCTION

Since the recent publication of its genome sequence (IBI, 2010), Brachypodium distachyon has become increasingly investigated as a model plant for temperate cereals, forage grasses and biofuel grass crops (Mur et al., 2011, and references therein). The choice of this annual species was based on several crucial attributes that made it an optimal model for advanced plant breeding programmes, such as: (1) the possession of one of the smallest monocot and grass genomes (1C approx. 0·3 pg, Wolny and Hasterok, 2009), comprising mostly single- or low-copy repetitive DNA (IBI, 2010); (2) its short life cycle, with a minimum of a 6 weeks from seed to seed (Garvin et al., 2008; Mur et al., 2011); (3) its self-fertility that ensures the generation of pure inbred lines within two generation cycles (Vogel et al., 2009); (5) its high germination rates both in the wild and under controlled greenhouse conditions (Garvin et al., 2008; Vogel et al., 2009); and (5) its closer relationship (Catalán et al., 1997; GPWG, 2001) to the economically important Triticeae species (Triticum spp., Hordeum spp.) than Oryza sativa, the tropical cereal with a fully sequenced genome, which gives it greater relevance in the development of a wide range of comparative analyses of gene expression and regulatory mechanisms (Mur et al., 2011).

Brachypodium distachyon is an annual pooid grass of relatively small stature that occurs naturally in the circum-Mediterranean region (from the Macaronesian islands to central Asia, and from southern Europe to northern Africa and Ethiopia) where it grows in xeric to mesic ephemeral pastures and open woodlands, between sea level and slightly above 2000 m altitude (Schippmann, 1991; Garvin et al., 2008; Vogel et al., 2009). It has been introduced in disturbed areas of central Europe, North and South America, Australia and South Africa (Schippmann, 1991; Garvin et al., 2008; Mur et al., 2011). A detailed description and synonymy of the species were provided by Schippmann (1991) in his comprehensive review of the European Brachypodium. This author recorded up to 37 heterotypic validly described synonyms for B. distachyon. Since Link (1827), numerous authors placed B. distachyon in Trachynia Link, a monotypic genus separating this annual species from the remaining perennial Brachypodium species. However, the lack of other diagnostic traits and the overall congruence of the shared characters with the perennial species prompted recent authors to classify the annual species within the broad Brachypodium genus (Schippmann, 1991, and references therein; Catalán et al., 1995). Brachypodium distachyon (basionym Bromus distachyos L., Flora Palaest. 13, 1756) was neotypified by Schippmann and Jarvis (1988) with the specimen LINN no. 93·48. Some of the synonymized names were described as annual species independent from B. distachyon (e.g. B. megastachyum Besser ex Schult. & J. J. Schult., B. macrostachyum Besser ex Schult. & J. J. Schult.). According to Schippmann (1991) their features fall within the large variation range of the highly variable B. distachyon and they should be regarded as synonyms of this species.

The first evolutionary analyses of Brachypodium indicated that the circum-Mediterranean B. distachyon was an early branching lineage within the genus, surprisingly close to the American short-rhizomatous perennial B. mexicanum. The two taxa were believed to share a base chromosome number of x = 5, which is different from the base chromosome numbers of x = 7, 8 and 9 of the other 12–14 Old World long-rhizomatous perennials (Shi et al., 1993; Catalán et al., 1995; Catalán and Olmstead, 2000). Whereas Catalán and Olmstead (2000) favoured an early split of the annual B. distachyon based on plastid and nuclear rDNA sequence data, followed by the consecutive divergence of B. mexicanum and the remaining members of the core perennials, Catalán et al. (1995) using nuclear randomly amplified polymorphic DNA data could not unambiguously differentiate between the early-branching of a B. distachyon/B. mexicanum clade and the recently evolved core perennial clade. A recent survey by Wolny et al. (2011), based on several single-copy nuclear genes, clarified the relationship between various Brachypodium species and revealed the potential hybrid nature of some taxa, including the sequenced B. distachyon accession.

The first karyological analyses of B. distachyon showed three different putative ploidy levels (2n = 10, 20 and 30) (Robertson, 1981). The three cytotypes have been traditionally regarded as constituting an autopolyploid series, derived from an initial diploid race with a base chromosome number of x = 5 and 2n = 10 chromosomes that subsequently generated the tetraploid and hexaploid derivatives, with 2n = 20 and 2n = 30 chromosomes, all sharing the same base chromosome number of x = 5 (Talavera, 1978; Robertson, 1981). However, modern comparative cytogenetic studies based on fluorescence in situ hybridization (FISH) with total genomic DNA (GISH), as well as FISH with ribosomal DNA multicopy genes and ‘single-locus’ BAC (bacterial artificial chromosome) clones indicated that the 2n = 10 and 2n = 20 chromosome races in fact correspond to two different diploid taxa, each with a distinct base chromosome number of x = 5 and x = 10, respectively, and the 2n = 30 race corresponding to their derived allotetraploid (Hasterok et al., 2004, 2006). This has recently been corroborated by Idziak et al. (2011) using comparative chromosome painting (CCP), which represents one of the most advanced and informative cytomolecular techniques. In spite of this, most authors still refer to the three cytotypes as part of a single species or single autopolyploid series (Shang et al., 2011; http://www.brachypodium.org/).

The International Brachypodium Initiative (http://www.brachypodium.org/) set the foundations for an exhaustive exploration of the B. distachyon genome and for comparative genomics with other grasses of interest, that ultimately yielded the 8× coverage of the fully sequenced genome (IBI, 2010). This was conducted with a B. distachyon Bd21 diploid inbred line with 2n = 10 chromosomes derived from individuals originating from its easternmost native distribution area in the Mediterranean – south-west Asian region (Iraq; Vogel et al., 2009; IBI, 2010). At the time of writing, a new project has been undertaken to re-sequence another 50 B. distachyon 2n = 10 inbred lines derived from individuals originating in both the easternmost and westernmost parts of the Mediterranean region (http://brachypodium.pw.usda.gov/). Additionally, there are proposals to sequence the genomes of the other B. distachyon sensu lato (s.l.) 2n = 20 and 2n = 30 cytotypes (J. P. Vogel, USDA, Albany, CA, pers. comm.).

Due to the current uncertainty about the taxonomic identity and evolution of the different B. distachyon s.l. cytotypes, and because of their importance as model grasses for temperate diploid and allopolyploid cereals and for comparative phylogenomics, we have conducted a comprehensive systematic study of the B. distachyon complex using multidisciplinary morphometric, cytogenetic, phylogenetic and taxonomic analyses. Our study aims to confirm: (1) the existence of statistically significant morphoanatomical differences between the three cytotypes through the study of both greenhouse-propagated lines and wild individuals; (2) the cytogenetic distinctiveness of each cytotype through a comparative study of rDNA-FISH-, GISH-, BAC-FISH- and CCP-based mapping; (3) the evolutionary relationships of the three cytotypes and other selected perennial Brachypodium species through a phylogenetic study based on plastid and nuclear genes; (4) the genomic substitution rates of the B. distachyon complex cytotypes and those of other Brachypodium species through comparative evolutionary rate tests; (5) the B. distachyon type and the International Code of Botanical Nomenclature (ICBN) rules that would support the description of two novel species for the 2n = 20 and 2n = 30 cytotypes and the selection of a new epitype for B. distachyon 2n = 10.

MATERIALS AND METHODS

Sampling

The study was based on the Brachypodium distachyon (L.) P.Beauv. s.l. inbred lines generated for the three cytotypes: 2n = 10: Bd21 (fully sequenced genome), ABR1; 2n = 20: ABR114; 2n = 30: ABR110, ABR113, ABR117, which were cultivated at Aberystwyth (ABR) University (Jenkins et al., 2003; Mur et al., 2011) (Table 1, Fig. 1). Additionally, new cytotypic 2n = 10, 20 and 30 samples, derived from seeds and from new germplasm collections and characterized by chromosome counting or flow cytometry measurements, were also included in the analyses, totalling 335 individuals from 36 lines or populations (Table 1). Within the cultivated plants, 216 individuals corresponding to 12 inbred lines of the three cytotypes were used in the basic statistical morphometric study, while these samples plus 119 individuals from 24 wild populations and the B. distachyon type specimen were used in a discriminant analysis validation study (see below). Different subsets of the cultivated and wild B. distachyon s.l. samples plus 28 additional Brachypodium and grass samples were used in the respective cytogenetic and phylogenetic studies (Table 1). To frame the phylogenetic analyses within a broad generic-level analysis and to conduct comparative substitution rate analyses, other Brachypodium representatives were also incorporated into the evolutionary study. The sampling included the short-rhizomatous perennial B. mexicanum (x = 5 or x = 10), and other long-rhizomatous species thought to belong to the core perennial clade (x = 7, 8, 9): B. arbuscula, B. boissieri, B. glaucovirens, B. phoenicoides, B. pinnatum, B. retusum, B. rupestre and B. sylvaticum (Table 1). Samples of core-pooids Triticeae (Hordeum vulgare, Secale cereale, Triticum aestivum) and Loliinae (Festuca pratensis, Lolium perenne) and of basal-pooids Meliceae (Melica ciliata, Glyceria declinata) were used as outgroups in the phylogenetic analyses.

Table 1.

List of Brachypodium and outgroup samples used in the phenetic, cytogenetic and phylogenetic analyses

Taxon	Population/Line code	Locality	Chromosomes	Ploidy	N
B. distachyon	Bdistachyon21	Iraq: Bd21 USDA genome sequenced	2n = 10	2x	18
B. distachyon	Bdistachyon1	Turkey: Kiresehir, Kaman, ABR1	2n = 10	2x	13
B. distachyon	Bdistachyon306	France: Herault, Octon	2n = 10	2x	1
B. distachyon	Bdistachyon384	Slovenia: Lubjana	2n = 10	2x	1
B. distachyon	Bdistachyon400	Spain: Huesca, Ibieca, San Miguel de Foces	2n = 10	2x	1
B. distachyon	Bdistachyon401	Spain: Huesca, Jaca, Guasillo	2n = 10	2x	1
B. distachyon	Bdistachyon5	Spain: Huesca, Sariñena	2n = 10	2x	5
B. distachyon	Bdistachyon12	Spain: Huesca, Jaca, Banaguas	2n = 10	2x	6
B. distachyon	Bdistachyon14	Spain: Huesca, Puente de la Reina	2n = 10	2x	5
B. distachyon	Bdistachyon15	Spain: Zaragoza, Murillo de Gallego	2n = 10	2x	6
B. distachyon	Bdistachyon18	Spain: Huesca, Arens	2n = 10	2x	5
B. distachyon	Bdistachyon19	Spain: Huesca, Adahuesca, San Jerónimo	2n = 10	2x	4
B. distachyon	Bdistachyon20	Spain: Huesca, Barbastro	2n = 10	2x	5
B. distachyon	Bdistachyon21	Spain: Huesca, Abizanda	2n = 10	2x	5
B. distachyon	Bdistachyon22	Spain: Huesca, Berdun	2n = 10	2x	7
B. distachyon	Bdistachyon23	Spain: Zaragoza, Miramont	2n = 10	2x	7
B. distachyon	Bdistachyon 24	Spain: Zaragoza, Sigüés	2n = 10	2x	7
B. distachyon	Bdistachyon25	Spain: Navarra, Foz de Lumbier	2n = 10	2x	6
B. distachyon	Bdistachyon27	Spain: Navarra, Puerto del Perdon	2n = 10	2x	6
B. distachyon	Bdistachyon28	Spain: Navarra, Los Arcos	2n = 10	2x	5
B. distachyon	Bdistachyon30	Spain: Lleida, Fondedou	2n = 10	2x	4
B. distachyon	Bdistachyon31	Spain: Lleida, Castillo de Mur	2n = 10	2x	5
B. distachyon	Bdistachyon32	Spain: Lleida, Les Pallagues	2n = 10	2x	6
B. distachyon	Bdistachyon41	Spain: Teruel, Calaceite	2n = 10	2x	5
B. distachyon	Bdistachyon42	Spain: Zaragoza, Belchite	2n = 10	2x	3
B. distachyon	BdistachyonLINN	Unknown locality: B. distachyon LINN 93·48 type	?	?	1
B. stacei	Bdistachyon114	Spain: Balearic isles, Formentera, ABR114	2n = 20	2x	73
B. stacei	Bdistachyon383	France: Corsica, Bonifacio	2n = 20	2x	1
B. stacei	Bdistachyon385	Portugal: Lisboa 401	2n = 20	2x	1
B. cf. stacei	BdistachyonGB	USA: R. Riggins, Calif, Genbank L11578	?	?	1
B. hybridum	Bdistachyon100	Iran: Kalafabad, ABR100	2n = 30	4x	37
B. hybridum	Bdistachyon110	France: Aude, ABR110	2n = 30	4x	1
B. hybridum	Bdistachyon113	Portugal: Lisboa, ABR113	2n = 30	4x	22
B. hybridum	Bdistachyon117	Afghanistan: PI219965, ABR117	2n = 30	4x	17
B. hybridum	Bdistachyon101	S. Africa: near Darling, Cape Province, ABR101	2n = 30	4x	6
B. hybridum	Bdistachyon105	Morocco: near Ongada, ABR105	2n = 30	4x	6
B. hybridum	Bdistachyon112	Belgium: Corse, Leigem, ABR112	2n = 30	4x	6
B. hybridum	Bdistachyon116	Afghanistan: PI219961, ABR116	2n = 30	4x	6
B. hybridum	Bdistachyon121	Iran: PI226452, ABR121	2n = 30	4x	6
B. hybridum	Bdistachyon137	Australia: West Australia, PI533015, ABR137	2n = 30	4x	6
B. hybridum	Bdistachyon402	Spain: Zaragoza, La Alfranca	2n = 30	4x	1
B. hybridum	Bdistachyon403	Spain: Girona, Cadaques	2n = 30	4x	1
B. hybridum	Bdistachyon16	Spain: Huesca, Graus	2n = 30	4x	4
B. hybridum	Bdistachyon36	Spain: Girona, Cap de Lladro	2n = 30	4x	4
B. hybridum	Bdistachyon37	Spain: Barcelona, Montjuich	2n = 30	4x	4
B. hybridum	Bdistachyon39	Spain: Tarragona, Amposta	2n = 30	4x	4
B. hybridum	Bdistachyon40	Spain: Tarragona, Poble Nou del Delta	2n = 30	4x	5
B. arbuscula	Barbuscula1	Spain: Canary isles, Gomera, Barb500	2n = 18	2x	1
B. arbuscula	Barbuscula2	Spain: Canary isles, Gran Canaria	2n = 18	2x	1
B. boissieri	B boissieri	Spain: Granada, Sierra Nevada	cf. 2n = 42, 46	8x	1
B. glaucovirens	Bglaucovirens	Greece: 4202	2n = 16	2x	1
B. mexicanum	Bmexicanum	Mexico: Hidalgo, Sierra de Pachuca, Bmex347	2n = 40	4x or 8x	1
B. phoenicoides	Bphoenicoides1	France: Var, Montferrat, Bpho39	2n = 28	4x	1
B. phoenicoides	Bphoenicoides2	Portugal: PI 283194, Bph	2n = 28	4x	1
B. pinnatum	Bpinnatum1	England: Sussex: Fairlight, Bpin8	2n = 28	4x	1
B. pinnatum	Bpinnatum2	Norway: PI 345982, Bpm	2n = 18	2x	1
B. pinnatum	Bpinnatum3	Turkey: PI251445, Bpk	2n = 28	4x	1
B. retusum	Bretusum1	Spain: Huesca, Fraga	cf. 2n = 36	4x	1
B. retusum	Bretusum2	Greece: 4195	2n = 38	4x	1
B. rupestre	Brupestre1	Spain: Navarra, Betelu, PC2393	cf. 2n = 28	4x	1
B. rupestre	Brupestre2	Russia: PI 440172 (1332), Brv	2n = 18	2x	1
B. rupestre	Brupestre3	Greece: 4196	2n = 18	2x	1
B. sylvaticum	Bsylvaticum1	Hungary: Balaton, Bsyl131	2n = 18	2x	1
B. sylvaticum	Bsylvaticum2	Australia: PI 297868, Bsa	2n = 18	2x	1
Outgroups
Melica ciliata	Melica ciliata	Spain: Huesca: Fraga	2n = 18	2x	1
Glyceria declinata	Glyceria declinata	Spain: Cantabria: Picos de Europa	2n = 20	2x	1
Secale cereale	Secale cereale	GenBank	2n = 14	2x	1
Hordeum vulgare	Hordeum vulgare	GenBank	2n = 14	2x	1
Triticum aestivum	Triticum aestivum	GenBank	2n = 42	6x	1
Festuca pratensis	Festuca pratensis	England: Wilshire, Calne, and GenBank	2n = 14	2x	1
Lolium perenne	Lolium perenne	England: Leicester, and GenBank	2n = 14	2x	1

Chromosome numbers of Brachypodium spp. samples were obtained from chromosome counts and flow cytometry analysis. N, number of studied individuals per population or inbred line used in the phenetic analysis. Different propagated individuals from cultivated inbred lines were used in the cytogenetic analysis. Different single individual samples per population or inbred line were used in the phylogenetic analysis.

Authorities for Brachypodium spp. names: B. distachyon (L.) P. Beauv., B. stacei Catalán, Joch. Müll., Mur & Langdon sp. nov., B. hybridum Catalán, Joch. Müll., Hasterok & Jenkins sp. nov., B. arbuscula Gay ex Knoche, B. boissieri Nym., B. glaucovirens (Murb.) Sagorski (syn. B. sylvaticum subsp. glaucovirens (St-Yves) Murb.), B. mexicanum (Roem. & Schult.) Link, B. phoenicoides (L.) P. Beauv. ex Roem. & Schult., B. pinnatum (L.) P. Beuv., B. retusum (Pers.) P. Beauv., B. rupestre (Host) Roem. & Schult., B. sylvaticum (Huds.) P. Beauv.

Fig. 1.

Descriptive images of the three Brachypodium distachyon s.l. cytotypes, grown under standard greenhouse conditions (see text) until maturity. (A–C) Habit and (D–F) flowering spikelets: (A, D) B. distachyon 2n = 10 (Bdistachyon21); (B, E) B. distachyon 2n = 20 (Bdistachyon114); (C, F) B. distachyon 2n = 30 (Bdistachyon113). Scale in centimetres.

Morphometric analysis

The taxonomic variability described within B. distachyon s.l. (Schippmann, 1991; Vogel et al., 2009) and in other diploid-to-polyploid grass species (Metcalfe, 1960) allowed us to select 15 potentially informative morphoanatomical traits that could be used to separate and identify the three B. distachyon cytotypes (Table 2). Twelve of them were quantitative; (plant) height (H), second leaf length (second leaf from the base of the plant, SLL), second leaf width (SLW), (stomata) leaf guard cell length (LGCL), inflorescence length (IL), spikelet length (total length, without awns; SLa), spikelet length (from the base to the apex of the fourth lemma, without awns; SLb), upper glume length (UGL), lemma length (from the basal floret, LL), awn length (the longest within the spikelet, AL), caryopsis length (from the basal floret, CL) and pollen grain length (PGL). A further three discrete characters were used: number of nodes of tallest culm (NNTC), number of spikelets per inflorescence (NSI) and number of flowers per inflorescence (NFI).

Table 2.

(A) Statistical descriptors of 15 morphoanatomical characters analysed in greenhouse-propagated individuals of the three Brachypodium distachyon s.l. cytotypes: 2n = 10: Bdistachyon21 and Bdistachyon1; 2n = 20: Bdistachyon114; 2n = 30: Bdistachyon110, Bdistachyon113, Bdistachyon117 (plus Bdistachyon101, Bdistachyon105, Bdistachyon112, Bdistachyon116, Bdistachyon121, Bdistachyon 137 for character CL)

Cytotype	LGCL (µm)	PGL (µm)	H (cm)	NNTC	SLL (cm)	SLW (mm)	IL (cm)	NSI	SLA (mm)	SLB (mm)	NFI	UGL (mm)	LL (mm)	AL (mm)	CL (mm)
2n = 10
N	31	31	8	10	10	10	13	13	13	13	12	13	11	13	12
Max.	29·00	35·00	35·00	5·00	8·50	3·30	4·80	4·00	22·00	14·60	9·00	8·20	9·00	13·00	7·20
Min.	18·00	27·00	18·00	2·00	4·30	2·50	1·50	1·00	12·50	12·00	5·00	6·20	7·10	9·50	6·20
Mean	23·26	29·87	26·13	3·30	6·68	2·84	3·25	2·69	16·31	12·70	7·00	7·23	8·05	11·45	6·75
s.d.	2·54	1·84	5·64	0·95	1·29	0·26	1·08	0·95	3·11	0·71	1·48	0·60	0·61	1·16	0·33
Variance	6·46	3·38	31·84	0·90	1·67	0·07	1·17	0·90	9·65	0·50	2·18	0·36	0·37	1·34	0·11
2n = 20
N	73	38	9	16	14	13	13	12	21	22	21	21	22	21	12
Max.	38·00	38·00	46·00	4·00	14·00	7·00	9·50	4·00	41·00	30·00	14·00	7·60	11·20	10·20	6·30
Min.	20·00	26·00	26·00	2·00	7·00	2·60	2·10	1·00	15·00	13·00	6·00	5·00	7·40	2·30	4·80
Mean	28·22	32·58	35·56	2·94	11·06	4·15	6·12	2·75	22·88	16·11	9·24	6·24	8·91	7·27	5·68
s.d.	3·88	3·17	6·77	0·77	1·80	1·42	2·13	0·87	5·80	3·31	2·45	0·80	0·86	2·33	0·56
Variance	15·03	10·03	45·78	0·60	3·23	2·02	4·53	0·75	33·70	10·93	5·99	0·65	0·74	5·41	0·31
2n = 30
N	75	56	11	18	15	17	20	21	20	20	20	20	19	18	55
Max.	40·00	48·00	48·00	7·00	16·00	4·00	5·50	5·00	38·00	18·00	16·00	9·70	12·40	12·20	8·20
Min.	28·00	34·00	17·00	1·00	6·00	1·10	2·00	1·00	15·00	13·80	5·00	7·00	9·00	8·00	5·70
Mean	33·64	38·27	33·09	3·72	10·30	2·75	3·93	2·71	22·17	15·56	8·15	7·92	10·49	9·68	7·02
s.d.	2·25	3·08	11·26	1·49	2·80	0·85	1·05	1·31	6·44	1·13	3·17	0·72	0·96	1·26	0·60
Variance	5·07	9·51	126·89	2·21	7·85	0·73	1·10	1·71	41·49	1·27	10·03	0·52	0·93	1·59	0·36

N, number of analysed individuals; s.d., standard deviation. See text for definitions of abbreviations of variables.

(B) Significance tests of mean values of the 15 analysed characters

	Significantly discriminate 2n = 10 vs. 2n = 20 vs. 2n = 30 cytotypes					Significantly discriminate 2n = 10 vs. 2n = 20 + 2n = 30 cytotypes			Significantly discriminate 2n = 10 + 2n = 30 vs. 2n = 20 cytotypes			Do not discriminate among cytotypes
	LGCL (µm)	PGL (µm)	UGL (mm)	LL (mm)	AL (mm)	SLL (cm)	SLA (mm)	SLB (mm)	SLW (mm)	IL (cm)	CL (mm)	NNTC	H (cm)	NFI	NSI
2n = 10	23·2  ±  2·5c	29·9  ±  1·8c	7·2  ±  0·6b	8·1  ±  0·6c	11·4  ±  1·1a	6·7  ±  1·3b	16·3  ±  3·1b	12·7  ±  0·7b	2·8  ±  0·3b	3·2  ±  1·1a	6·7  ±  0·3a	3·3  ±  0·9a	26·1  ±  5·6b	7·0  ±  1·5b	2·7  ±  0·9a
2n = 20	28·2  ±  3·9b	32·6  ±  3·2b	6·2  ±  0·8c	8·9  ±  0·9b	7·3  ±  2·3c	11·1  ±  1·8a	22·9  ±  5·8a	16·1  ±  3·3a	4·1  ±  1·4a	6·1  ±  2·1b	5·7  ±  0·6b	2·9  ±  0·8a	35·5  ±  6·8a	9·2  ±  2·4a	2·7  ±  0·9a
2n = 30	33·6  ±  2·2a	38·3  ±  3·1a	7·9  ±  0·7a	10·5  ±  1a	9·7  ±  1·3b	10·3  ±  2·8a	22·2  ±  6·4a	15·5  ±  1·1a	2·7  ±  0·8b	3·9  ±  1·0a	7·0  ±  0·6a	3·7  ±  1·5a	33·1  ±  11·3ab	8·1  ±  3·2ab	2·7  ±  1·3a
H/χ2	110·3	82·3	28·4	31·2	29·2	17·9	14·3	26·7	11·9	15·7	27·9	3·41	5·2	6·6	0·02
d.f.												2	2	2	2
P	<0·001	<0·001	<0·001	<0·001	<0·001	<0·001	<0·001	<0·001	<0·002	<0·001	<0·001	0·18	0·07	0·03	0·98

Mean  ±  s.d. and ANOVA χ²or Kruskal–Wallis H tests of variables used for comparisons among cytotypes (d.f. 2). Superscripts denote Mann–Whitney pairwise comparisons between cytotypes; means with the same letter do not differ significantly (P < 0·05). For abbreviations of variables see text.

Individuals propagated from the cytogenetically characterized inbred germplasm lines Bd21 and ABR1 (2n = 10), ABR114 (2n = 20), and ABR110, ABR113 and ABR117 (2n = 30) were cultivated under standard greenhouse conditions (Table 1). Plants were grown under a 16-h light period at 20 °C, illuminated with 55-W high-frequency lighting tubes (4580-Lumen output) supplemented with 2× 30-W clear tube cooled lighting and placed between 60 and 100 cm of the light bank with light intensities always exceeding 100 mol m⁻² s⁻¹. Individual plants were grown routinely on compost supplemented with gravel to approx. 50 % (v/v) to improve drainage. Plants were usually watered at 2-d intervals and never allowed to stand in water. The individuals were screened and measured for the above mentioned 15 characters. Individuals from six other 2n = 30 lines (ABR101, ABR105, ABR112, ABR116, ABR121, ABR137) were also measured for character CL (Table 1). The standardized cultivation conditions guaranteed the measurement of genetically fixed features within the studied individuals from the lines belonging to the three cytotypes. Simple statistical descriptors of the intra- and inter-cytotypic phenetic diversity (mean, range, s.d., box plots of median, range and percentiles) were calculated from the data. Inter-cytotype response variables were estimated through one-way ANOVA chi-square tests or through non-parametric Kruskal–Wallis and pairwise Mann–Whitney tests when the variables complied or not, respectively, with requirements of normality.

To test if the significant phenetic differences representative of the three cytotypes observed among the greenhouse-propagated individuals (see Results) were also present in wild individuals, a biological field validation approach was conducted through multivariate classification discriminant analysis (DA) on a set of 102 individuals from 19 wild populations with 2n = 10 chromosomes and 17 individuals from five wild populations with 2n = 30 chromosomes from Spain (Table 1). This data set included the 2n = 10 (Bd21, ABR1), 2n = 20 (ABR114) and 2n = 30 (ABR110, ABR113, ABR117) cultivated individuals as reference samples. Due to the extreme rarity of wild populations of the 2n = 20 cytotype, only propagated individuals from the ABR114 line were included in this analysis. From these data, and due to incomplete sampling of some characters in all individuals, we calculated the average value of each character for each line or population across their respective studied individuals (Table 1) and used the averaged line-population samples (36) to perform the discriminant analysis. DA estimated the probability of membership of each wild sample to cytotaxonomically predefined groups (2n = 10, 20 and 30), which were characterized by the phenetically different propagated lines (see Results), allowing the identification of the more discriminating variables by means of Fisher's coefficient (Fisher, 1936; Anderson, 1996) at the significant threshold value of 0·05. The posterior probability of classification of each sample and the Wilks' Lambda value of each discriminant function were calculated (Wilks, 1932). A Wilks' Lambda value closer to zero indicated a better discrimination between the predefined groups.

The neotype of B. distachyon (LINN 93·48) was also studied to infer its potential membership to any of the three morphometric cytotypes. This sample was subjected to a restricted DA with a subset of the characters that could be measured on the type specimen (H, SLL, SLW, IL, SLb, UGL, LL, AL, NNTC, NSI, NFI). All statistical analyses were conducted using SPSS 19·0.

Cytogenetic analyses

Mitotic chromosome preparations were made as described in detail in Jenkins and Hasterok (2007). In brief, the seeds were germinated in the dark in Petri dishes on filter paper moistened with tap water at 22 °C. When roots reached a length of 1·5–2·0 cm, whole seedlings were immersed in ice-cold water followed by incubation for 24 h and fixation in 3 : 1 (v/v) methanol/glacial acetic acid at room temperature for several hours, and then stored at –20 °C until required. Excised roots were washed in a few changes of 0·01 m citric acid – sodium citrate buffer (pH 4·8) and digested in a mixture of enzymes comprising 20 % (v/v) pectinase, 1 % (w/v) cellulase and 1 % (w/v) cellulase ‘Onozuka R-10’ for 2 h at 37 °C. Meristems were dissected out from root tips and squashed in a drop of 45 % acetic acid. After freezing, coverslips were removed and the preparations were briefly post-fixed in a pre-chilled (–20 °C) 3 : 1 ethanol/glacial acetic acid mix followed by dehydration in absolute ethanol and air drying. For meiotic chromosomes, immature inflorescences were collected, fixed in fresh 3 : 1 absolute methanol/glacial acetic acid mixture for 3× 24 h at room temperature and stored at –20 °C until required. Anthers were isolated and washed in 10 mm citric acid – sodium citrate buffer and digested enzymatically for 2 h at 37 °C in a mixture comprising 10 % (v/v) pectinase, 0·65 % (w/v) cellulase Onozuka R-10, 0·5 % (w/v) cellulase, 0·15 % (w/v) cytohelicase and 0·15 % (w/v) pectolyase in 10 mm citric acid – sodium citrate buffer. Four to six anthers were squashed in drops of 45 % acetic acid, frozen and air dried.

Probe templates for GISH were derived from total genomic DNA of the cytotypic lines ABR1 (2n = 10), ABR114 (2n = 20) and ABR113 (2n = 30) (Hasterok et al., 2004). The templates for FISH with ribosomal DNA were obtained from a 2·3-kb ClaI sub-clone of the 25S rDNA coding region of Aradopsis thaliana (Unfried and Gruendler, 1990) and from a pTa794 clone of the 5S rDNA coding region of wheat (Gerlach and Dyer, 1980). The probes specific for the short and long arm, respectively, of chromosome 2 of B. distachyon (2n = 10) were based on two BAC clones, ABR1-41-E10 and ABR5-1-H3, derived from the large-insert genomic library of this species made by Hasterok et al. (2006). The probes were made by nick translation or PCR labelling (5S rDNA) as described elsewhere (Hasterok et al., 2004; Jenkins and Hasterok, 2007) using either tetramethyl-rhodamine-5-dUTP or digoxigenin-11-dUTP. For CCP the pools of BAC clones for the short and long arm of B. distachyon chromosome 5 (Bd5) were used. BACs containing more than 30 % repetitive sequences were removed from the pools. The probes were labelled with digoxigenin-11-dUTP for short chromosome arms and with tetramethyl-rhodamine-5-dUTP for long chromosome arms as described in Idziak et al. (2011). The list of BACs may be requested from the authors.

FISH was carried out as described in detail previously (Hasterok et al., 2004, 2006; Jenkins and Hasterok; 2007, Idziak et al., 2011). The standard hybridization mixture consisted of 50 % deionized formamide, 2× saline sodium citrate (SSC), 10 % (w/v) dextran sulphate, 10 µg μL⁻¹ sonicated salmon sperm DNA and 0·5 % sodium dodecylsulphate. For cross-species (heterologous) BAC-FISH and CCP, the formamide concentration was reduced to 30 %. All probes were mixed to a final concentration each of about 2·5 ng μL⁻¹. Probes were denatured separately in hybridization mixtures (80 °C for 10 min), applied to the preparations and denatured together (70 °C for 4·5 min). The slides with chromosome material were allowed to hybridize with the probes for 12–20 h in a humid chamber at 37 °C. For CCP the renaturation time was extended to 45–55 h. Post-hybridization washes were routinely carried out for 10 min in 20 % deionized formamide in 2× SSC at 37 °C with the exception of cross-species BAC-FISH (10 min in 10 % deionized formamide in 0·1 % SSC at 42 °C). Immunodetection of digoxygenated probes was performed according to standard protocols using fluorescein isothiocyanate-conjugated antidigoxigenin antibodies. Finally, the preparations were mounted and counterstained in Vectashield containing 2·5 µg mL⁻¹ 4′,6-diamidino-2-phenylindole (DAPI). Photomicrographs were taken using monochromatic or colour CCD cameras attached to epifluorescence microscopes using respective narrow band filters. All images were processed uniformly and superimposed using relevant software.

Flow cytometry procedures were carried out as described in Hajdera et al. (2003). In brief, suspensions of nuclei were prepared from young leaves and stained with PI (propidium iodide) for DNA content measurement using a DAKO Galaxy flow cytometer equipped with an air-cooled argon ion laser. Nuclei from young leaves of Brassica rapa ‘Goldball’ (0·97 pg/2C DNA; D. Siwinska, unpubl. res.) were used to calculate nuclear DNA content of B. distachyon 2n = 10 (Wolny and Hasterok, 2009), while in the case of B. distachyon 2n = 20 and B. distachyon 2n = 30, Brassica rapa ‘Goldball’ (0·97 pg/2C DNA) and Lycopersicon esculentum ‘Stupicke’ (1·96 pg/2C DNA; Dolezel et al., 1992) were used as standards, respectively. The data were processed using FloMax (Partec GmbH, Münster, Germany).

Phylogenetic analysis, dating analysis and evolutionary rate analysis

Sequences from two plastid (ndhF, trnLF) and five nuclear (ITS, ETS, CAL, DGAT, GI) loci were used in the phylogenetic analyses of the B. distachyon s.l. cytotypes and selected representatives of Brachypodium (Table 1). The plastid data included the 3′-end coding region of the NAD dehydrogenase subunit F (ndhF) gene and the trnL(UAA) intron – trnL(UAA) exon – trnL(UAA)/trnF(GAA) spacer (trnLF) region that were amplified and sequenced using the procedures specified, respectively, in Catalán et al. (1997) and Torrecilla et al. (2003). The nuclear multicopy data included the consensus sequences of the internal transcribed spacer (ITS) and external transcribed spacer (ETS) of the ribosomal DNA repeat unit that were sequenced from PCR products following the procedures described by Hsiao et al. (1995) and Gillespie et al. (2009), respectively. Forward and reverse sequences of each sample were corrected and assembled using Sequencher 4·10·1. The nuclear single-copy gene data consisted of coding and intron regions of genes encoding a calmodulin-binding protein (CAL), diacyl glycerol acyl transferase (DGAT) and Gigantea (GI). CAL and GI were PCR amplified and sequenced as described in Wolny et al. (2011). DGAT was amplified with primers DGATE2F (GCAGTGAACAGCAGRCTCATTATTGAGAA) and DGATEWR (GTTGGCACTGAGAGTTTYAAGACTCTYTC); additional internal primers DGBRSEQF1 (CAACATCTGTCATCGTCTATCCAGTTG) and DGBRSEQR1A (GGTCAACTTTGACTATCCAGAAATGTG) were used for sequencing all species except B. distachyon 2n = 10 where DGBRSEQR1A was replaced by DGBRSEQR1B (GGTCAACTTTCACTATCCAGAAATGTG) and B. distachyon 2n = 20 where DGBRSEQR1A was replaced by DGBRSEQR1C (GGTCAACTTGGACTATCCAGAAATGTG) (DGBRSEQR1B and DGBRSEQR1C were also used as appropriate for B. distachyon 2n = 30 DGAT clones). PCR products from diploid species were sequenced directly; those from polyploid species were cloned and multiple clones were sequenced to recover consensus sequences for each allele. In total, 126 new Brachypodium and pooid sequences were generated in the present study [GenBank accession numbers JN187625–JN187649 (ndhF); JN187650–JN187674 (trnLF); JN187602–JN187624 (ITS); JN187580–JN187601 (ETS); JN589946–JN589947 (CAL); JN204845–JN204868 (DGAT); and JN589948–JN589952 (GI)]; these data were aligned with sequences obtained in our previous studies or retrieved from GenBank and used in the phylogenetic analyses (Table 1 and Supplementary Data Table S1, available online). Multiple sequence alignments of each separate data set were done using the Clustal algorithm option of MacClade 4·08 (Maddison and Maddison, 2005) or the Muscle server (Edgar, 2004; http://www.ebi.ac.uk/Tools/msa/muscle/) and adjusted manually. The final data sets consisted of 27 analysed samples/699 aligned positions for ndhF, 27/1037 for trnLF, 26/644 for ITS, 25/687 for ETS, 16 (21 sequences)/481 for CAL, 16 (27)/2543 for DGAT and 17 (27)/872 for GI. Potentially informative indels were scored as binary data for the ndhF (one indel), trnLF (six), DGAT (39) and GI (eight) data sets and used in the parsimony-based evolutionary analyses.

Phylogenetic analyses were performed through maximum parsimony (MP), maximum likelihood (ML) and Bayesian inference (BI) methods using the programs Paup* v.4·10b (Swofford, 2002), RAxML Blackbox server 7·2·8 (Stamatakis, 2006; http://phylobench.vital-it.ch/raxml-bb/index.php), which implements the search protocol of Stamatakis et al. (2008), and MrBayes 3·1·2 (Ronquist and Huelsenbeck, 2003), respectively. Independent analyses were conducted on each separate data matrix using close Meliceae or Triticeae outgroup representatives to root the trees. The MP analysis consisted of 10 000 replicates of random heuristic search with TBR (hold = 100, MulTrees = yes, nchuck = 100, and chuckscore = 10). MP trees were used to obtain a 50 % majority rule consensus tree. Branch support was computed through 10 000 bootstrap replicates under the above parsimony settings; but with ten repetitions of random addition (hold = 5 and nchuck = 5) per replicate. For ML analysis, a nucleotide substitution model was selected using the hierarchical likelihood ratio test and the Akaike criterion provided in MrModeltest 2·3 (Nylander, 2004). The ML analysis was conducted according to the GTR nucleotide substitution model leaving the program to estimate the parameters. Bootstrap support values were based on 100 rapid replicates using a GTRMIX model. BI was computed imposing the nst = 6 and rates = gamma substitution model to the nucleotide sequence partition. In total, 3750 Bayesian trees were obtained after performing two runs, each with 5000 000 generations and four chains, sampling trees every 1000 generations, and a burn-in option of 1250 trees per run once stability in the likelihood values was attained. The posterior probability values of branches in the Bayesian majority rule consensus tree of the 3750 trees were used as a measure of nodal support. Both plastid and nuclear rDNA combined data analyses were performed for those data sets with common sequenced accessions and a final combined analysis of plastid and nuclear rDNA data was performed in the search for a consensus species-tree that would recover the evolutionary history of the B. distachyon s.l. cytotypes and date their divergence.

The ages of splitting of the B. distachyon 2n = 10 and 2n = 20 diploid lineages and of other Brachypodium lineages were calculated from the combined plastid and nuclear rDNA data set using a Bayesian relaxed clock method implemented in Beast v.1·5·2 (Drummond and Rambaut, 2007). The inferred dates were derived from the reduced Bayesian tree depicted in Fig. 6H and secondary calibrations of 32 and 24 Ma for the estimated divergence of, respectively, Meliceae from the core pooids and Triticeae from Aveneae–Poeae (Vicentini et al., 2008). Data were analysed as partitioned (plastid vs. nuclear). Each partition was analysed using a GTR + Г model, with substitution models being unlinked across partitions. Best-fitting evolutionary models were selected according to the Akaike information criterion provided by the software MrModelTest v.2·3. The input files were composed with the software BEAUti v1·5·3 with the tree prior set as follows: ages for the divergence of Meliceae from the core pooids and Triticeae from Aveneae–Poeae of, respectively, 32 and 24 Ma (see references above); a lognormal prior distribution (logmean = 0; lognormal s.d. = 1·0; offsets set to 32 and 24 Ma) assigned to these calibration points following Ho and Phillips (2009): substitution rates for both partitions (plastid and nuclear) established according to Wolfe et al. (1987). The Yule process was chosen as speciation process for the two data sets. The Beast analysis consisted of a run of 40 million generations, with parameters and trees sampled every 2000 generations. One additional chain was run for 50 million generations to test for the influence of the priors on the posterior estimates. Log files were analysed with Tracer v1·5 (Rambaut and Drummond, 2007) to assess convergence and to ensure that the Markov chain Monte Carlo process had run long enough to get a valid estimate of the parameters. All resulting trees were used to compute a maximum clade credibility tree, with a burn-in of 10 %, using TreeAnnotator v1·5·3 (Drummond and Rambaut, 2007; Drummond et al., 2007)

Fig. 6.

Bayesian phylogenetic trees of Brachypodium representatives and outgroup species showing the relationships among the three Brachypodium distachyon s.l. cytotypes (2n = 10, 20, 30) and with respect to other congeners: (A) ndhF tree; (B) trnLF tree; (C) ITS tree; (D) ETS tree; (E) CAL tree; (F) DGAT tree; (G) GI tree; (H) combined cpDNA and rDNA tree. Values above branches indicate maximum parsimony bootstrap support (BS) >50 %/maximum likelihood BS/Bayesian inference posterior probability support (PPS). Blue, red and purple colours correspond, respectively, to the B. distachyon 2n = 10, 20 and 30 samples. Additional information on tree parameters is indicated in Supplementary Data S2, available online. Divergence times for the main Brachypodium lineages (estimates and intervals calculated using the Beast Bayesian relaxed clock method) are indicated in the nodes in the combined Bayesian tree in (H).

Due to the large mutational differences observed in both rDNA and plastid sequences among the B. distachyon s.l. cytotypes and with respect to the less variable perennial taxa, both absolute and relative substitution rates were estimated for each B. distachyon diploid cytotype (2n = 10, 2n = 20), with the aim to investigate the evolutionary processes involved in their biological diversification. The magnitude of variation of rate substitutions across their plastid and nuclear sequences were compared with those of the studied perennial taxa. To estimate the absolute evolutionary rates, pairwise genetic distances were computed for the separate nuclear rDNA and plastid data matrices based on the Tajima and Nei (1984) model with MEGA v.4 (Tamura et al., 2007). Substitution rates were calculated from the overall average of nucleotide substitutions (K) obtained from MEGA for each separate group and data set and the estimated time of divergence of the Brachypodium lineages from their most recent common ancestors (MRCAs) [D (in Ma), see Results] using the formula K/2 × D. Absolute substitution rates were also computed between different Brachypodium terminal tips of the nuclear rDNA and plastid trees. Relative rate tests (RRTs) were performed between the different Brachypodium species-pairs for the two genome regions analysed (rDNA, cpDNA) in the search for significant differences in nucleotide substitution rates among them. The RRTs were computed with HyPhy win3·2 (Kosakovsky-Pond et al., 2005) through distance-based likelihood ratio tests between pairs of sequences based on the Kimura (1980) two-parameter model using Secale cereale as outgroup.

RESULTS

Phenotypic diversity and differentiation of the three B. distachyon s.l. cytotypes

All 15 morphoanatomical characters showed both intra-cytotypic and inter-cytotypic diversity across the studied individuals (Table 2A, Fig. 2). Eleven of the 15 traits were useful in significantly discriminating: (1) all the three cytotypes from each other (LGCL, UGL, LL, AL, PGL), (2) the 2n = 10 vs. the 2n = 20 and 2n = 30 cytotypes (SLL, SLa, SLb) and (3) the 2n = 10 and 2n = 30 vs. the 2n = 20 cytotypes (SLW, IL, CL) (Table 2B). Only four characters did not significantly and unambiguously differentiate the cytotypes (H, NNTC, NFI, NSI). For most of the discriminating variables, the mean values tended to be higher in the allotetraploid 2n = 30 cytotype than in the diploid 2n = 10 and 2n = 20 cytotypes. LGCL (H = 110·3, P < 0·001 ), PGL (χ² = 82·3, P < 0·001) and LL (χ² = 31·2, P < 0·001) showed a significant increase from the lowest mean values of the 2n = 10 cytotype, through the intermediate ones of the 2n = 20 cytotype to the highest ones of the 2n = 30 cytotype. However, UGL (χ² = 28·4, P < 0·001) and CL (H = 27·9, P < 0·01) showed an increasing tendency from 2n = 20 through 2n = 10 to 2n = 30, IL (H = 15·7, P < 0·001) from 2n = 10 through 2n = 30 to 2n = 20, and AL (H = 29·2, P < 0·001) from 2n = 20 through 2n = 30 to 2n = 10. H (χ² = 5·2, P < 0·001), SLL (χ² = 17·9, P < 0·001), SLa (χ² = 14·3, P < 0·01) and SLb (H = 26·7, P < 0·001) showed lower mean values in 2n = 10 than in 2n = 20 and 2n = 30, and SLW (H = 11·9, P < 0·01) showed lower mean values in 2n = 10 and 2n = 30 than in 2n = 20 (Table 2A). The summarized box-plot descriptors reflected the non-overlapping percentile ranges of the most discriminating characters (LGCL, UGL, LL, AL, PGL; Fig. 2). These results clearly indicated that each B. distachyon cytotype could be phenotypically differentiated from the others and that these attributes were not subject to environmental influence but, most probably, were genetically fixed. The studied individuals of each cytotype line were consequently used as cytotypic reference samples in the classification discriminant analysis.

Fig. 2.

Box plots of simple statistics (median, percentiles, range) of 15 morphoanatomical characters analysed in greenhouse-propagated individuals (Table 1) of the three Brachypodium distachyon s.l. cytotypes: 2n = 10: Bdistachyon21 and Bdistachyon1; 2n = 20: Bdistachyon114; 2n = 30: Bdistachyon110, Bdistachyon113, Bdistachyon117 (plus Bdistachyon101, Bdistachyon105, Bdistachyon112, Bdistachyon116, Bdistachyon121 and Bdistachyon137 for character CL). (A) Variables that significantly discriminate 2n = 10 vs. 2n = 20 vs. 2n = 30; (B) 2n = 10 vs. 2n = 20 + 2n = 30 (C); 2n = 10 + 2n = 30 vs. 2n = 20; (D) variables that do not discriminate among cytotypes.

The DA validation approach incorporated the 216 propagated individual reference samples mentioned above for the 2n = 10, 2n = 20 and 2n = 30 cytotypes, plus 119 wild individual samples, cytotypically characterized by chromosome counts and flow cytometry as 2n = 10 or 2n = 30 individuals, with measurements averaged across individuals for each line-population sample (36 samples), and the cytotypically unknown B. distachyon LINN 93·48 type sample included as an uncertain sample (Table 1). The standard DA classification method based on the whole set of analysed characters (15) resulted in the correct classification of 100 % of the 2n = 10, 2n = 20 and 2n = 30 samples to their respective predefined groups. In the two-dimensional (2-D) DA scatterplot (Fig. 3A), the 2n = 10 and 2n = 20 samples vs. the 2n = 30 samples clustered, respectively, in the opposite sides of the first discriminant function, which accumulated 85·5 % of the total variation, whereas the 2n = 20 samples separated from the 2n = 10 samples along the second discriminant function, which accumulated 14·5 % of the variance. The DA confirmed that the morphological separation of the three cytotypes is supported by five characters (H, SLW, LGCL, SLb, UGL). Wilks' Lambda values of the first and second discriminant functions were, respectively, 0·038 and 0·390. The lowest Wilks' Lambda value obtained for the first discriminant function that separated the 2n = 10 and 2n = 20 vs. 2n = 30 cytotypes indicated the greater morphological differentiation of the allotetraploid cytotype from the diploids (see Results). However, the low Wilks' Lambda value obtained for the second discriminant function that separated the 2n = 10 and 2n = 20 cytotypes also supported their phenetic differentiation. The restricted DA that included the cytogenetically unknown B. distachyon LINN 93·48 type sample showed similar results (Fig. 3B). The 2n = 10, 20 and 30 samples clustered in the same spaces of the 2-D scatterplot although some 2n = 30 samples were more dispersed, probably due to their higher diversity and the lower resolution of the more limited data set, based on 11 characters and a separation supported by only four variables (H, SLW, SLb, UGL). In this analysis the LINN 93·48 sample was classified within the 2n = 20 group although it was spatially placed at an intermediate distance between the 2n = 20 and 2n = 10 groups (Fig. 3B), precluding any possible cytotypic assigment based on morphometry.

Fig. 3.

Two-dimensional scatterplots of classification DA of Brachypodium distachyon samples for taxonomic phenetic differentiation of the 2n = 10, 2n = 20 and 2n = 30 cytotypes. (A) Dataset of 15 morphometric characters analysed in 335 cultivated and wild individuals of the three cytotypes, with values averaged for each of the 36 line-population samples; (B) dataset of 11 morphometric characters analysed in the same set of 36 line-population samples plus the B. distachyon LINN 93·48 type sample (see text and Table 1). The first and second canonical discriminant functions explained 85·5 and 14·5 % of the intercytotypic taxonomic variation, respectively. 2n = 10, 2n = 20 and 2n = 30 as indicated in the key; open squares indicate the group centroid. LINN 93·48 is labelled in (B).

Cytogenetic identities of the B. distachyon s.l. cytotypes

Molecular cytogenetic analysis graphically illustrates the distinctiveness of the karyotypes of the three cytotypes. Brachypodium distachyon 2n = 10 (Bdistachyon1) was typical of the ten-chromosome diploid accessions, comprising a karyotype in which chromosomes 4 and 5 could be identified simply on the basis of 5S and 25S rDNA FISH probes (Fig. 4A). By contrast, B. distachyon 2n = 20 (Bdistachyon114) had 20 much smaller chromosomes which could not be the result of simply doubling the chromosome number of B. distachyon 2n = 10 as originally thought. Indeed, FISH with the same rDNA probes highlighted four chromosomes in this 2n = 20 cytotype (Fig. 4B), demonstrating that the two cytotypes were likely to be related to one another by Robertsonian fusion or fission events. The shapes, sizes and number of the chromosomes of B. distachyon 2n = 30 (Bdistachyon113) were consistent with its amphidiploid status, and confirmed that it was the product of a hybridization event between progenitors similar to B. distachyon 2n = 10 and 2n = 20. This was further confirmed by the pattern of the rDNA loci, which was a simple combination of those of the two diploid cytotypes (Fig. 4C). Simultaneous in situ hybridization with genomic DNA from B. distachyon 2n = 10 (Fig. 4D) and genomic DNA from B. distachyon 2n = 20 (Fig. 4E) clearly discriminated between the two chromosome sets in the allotetraploid (Fig. 4F). In an attempt to reconstruct at higher resolution the evolutionary relationships between the three cytotypes, single-locus BAC probes were systematically landed onto the chromosomes of the three karyotypes. BAC probes marking the long and short arms of chromosome 2 of B. distachyon 2n = 10 (Bdistachyon1) (Fig. 4G1, G2) identified two pairs of chromosomes in B. distachyon 2n = 20 (Bdistachyon114) (Fig. 4H1, H2), a result consistent with the above conclusion that the two cytotypes were related by structural rearrangements. As expected, these BACs identified three pairs of chromosomes in B. distachyon 2n = 30 (Bdistachyon113) (Fig. 4I1, I2), corroborating the evidence that it was an allotetraploid.

Fig. 4.

Patterns of cytogenetic markers on the somatic metaphase chromosomes of the three Brachypodium distachyon cytotypes (2n = 10: Bdistachyon1; 2n = 20: Bdistachyon114; 2n = 30: Bdistachyon113). FISH with 5S (red) and 25S (green) probes to B. distachyon 2n = 10 (A), B. distachyon 2n = 20 (B) and B. distachyon 2n = 30 (C). GISH with genomic DNA from B. distachyon 2n = 10 (D) and B. distachyon 2n = 20 (E) discriminates clearly between the two constituent chromosome sets of B. distachyon 2n = 30 (F). Single-locus BACs marking the long and short arms of chromosome 2 of B. distachyon 2n = 10 (G1) mark two and three pairs of chromosomes in B. distachyon 2n = 20 (H1) and B. distachyon 2n = 30 (I1), respectively. Ideograms of the marked chromosomes in the three cytotypes are shown in (G2), (H2) and (I2). Chromosomes are counterstained with DAPI (blue). Scale bars = 5 µm.

Chromosome painting with BAC clones originated from the short and long arm of B. distachyon chromosome 5 showed a strong signal along the whole bivalent length in B. distachyon 2n = 10 (Bdistachyon21) (Fig. 5A1). The same set of probes hybridized with chromosomes of B. distachyon 2n = 20 (Bdistachyon114) (Fig. 5B1) demonstrated the presence of one bivalent that was similar in size to its counterpart in the B. distachyon genome. The results observed in B. distachyon 2n = 30 (Bdistachyon113) (Fig. 5C1) represent the sum of what is seen in the two other cytotypes. The concise summary of chromosome painting is represented on ideograms (Fig. 5A2, C2).

Fig. 5.

Painting of Brachypodium distachyon chromosome 5 arm-specific pools of BAC clones onto first meiotic prophase chromosomes of B. distachyon 2n = 10 (Bdistachyon21) (A1), B. distachyon 2n = 20 (Bdistachyon114) (B1) and B. distachyon 2n = 30 (Bdistachyon113) (C1). Green and red signals mark short and long arms, respectively. The track of labelled bivalents is highlighted by white dotted lines. Red arrow indicates repetitive sequences. Painted bivalents are schematically shown in (A2–C2). Chromosomes are counterstained with DAPI (blue). Scale bars = 5 µm.

On the basis of chromosome sizes, shapes, relative numbers and patterns of various cytomolecular markers, the B. distachyon 2n = 10, 20 and 30 cytotypes could be considered as representing three distinctive species that are related by genome reorganization and hybridization.

Phylogenetic reconstruction, divergence times and comparative evolutionary rates of the B. distachyon s.l. cytotypes

Analysis of the maternally inherited plastid ndhF and trnLF sequences showed that the B. distachyon 2n = 20 and 2n = 30 cytotypic samples shared identical DNA sequences that were different from those of the 2n = 10 cytotypic samples. By contrast, analysis of the biparentally inherited nuclear rDNA ITS and ETS sequences, subjected to convergent evolution, showed that identical or close sequences were shared between the B. distachyon 2n = 10 and 2n = 30 cytotypic samples, which differed from those of the 2n = 20 cytotypic samples. Finally, a different resolution was obtained from analysis of the biparentally inherited but non-convergent single-copy nuclear CAL, DGAT and GI sequences, which showed that the B. distachyon 2n = 30 cytotypic sample (Bdistachyon113) contained alleles derived from both B. distachyon 2n = 20 (Bdistachyon114) and B. distachyon 2n = 10 (Bdistachyon21, Bdistachyon1) lineages. The number of synapomorphic nucleotide substitutions and indels detected within the B. distachyon s.l. cytotypes greatly exceed those found among the perennial representatives of Brachypodium in all plastid and nuclear data matrices. The variability observed within each data set is summarized in Supplementary Data Text, available online.

The phylogenetic reconstructions recovered from the MP, ML and BI analyses reflected the closeness of shared mutations observed in each data set. For each plastid and nuclear data matrix the three searches found an identical or congruent optimal tree (Fig. 6, Supplementary Data Text). The plastid ndhF and trnLF trees indicated that the B. distachyon 2n = 20 diploid cytotype was the maternal parent of the allotetraploid 2n = 30 cytotype (Fig. 6A, B), whereas the nuclear rDNA ITS and ETS trees showed that the B. distachyon diploid 2n = 10 cytotype was its other, i.e. paternal, parent (Fig. 6C, D). The nuclear CAL, DGAT and GI trees confirmed that the two diploid 2n = 20 and 2n = 10 cytotypes were probably components of the allotetraploid 2n = 30 cytotype (Fig. 6E–G).

The ndhF topology (Fig. 6A, Supplementary Data Text) showed a basal split of three relatively well-supported lineages; the clade of B. distachyon 2n = 20 and 2n = 30 sequences was resolved as sister to B. mexicanum, the clade of B. distachyon 2n = 10 sequences was resolved as sister to B. boissieri, but the sister relationship of the latter group to the strongly supported core perennial clade was poorly supported. Within the core perennial group, there was an early split of B. arbuscula followed by an unresolved polytomy of remaining taxa. The trnLF tree (Fig. 6B, Supplementary Data Text) recovered the successive divergence of B. mexicanum, the B. distachyon 2n = 20 + 2n = 30 clade, B. boissieri, the B. distachyon 2n = 10 clade and an unresolved clade of core perennials, although most of the basal and sub-basal splits were weakly supported. The combined analyses of concatenated ndhF and trnLF sequences recovered a consensus plastid topology (results not shown) that was coincident with that of the ndhF tree, but provided better support for the basal separation of the sister lineages that respectively included the B. distachyon 2n = 20 + 2n = 30 clade (sister to B. mexicanum) and the B. distachyon 2n = 10 clade (sister to B. boissieri).

The ITS tree (Fig. 6C, Supplementary Data Text) favoured a strongly supported early branching of B. boissieri and then a polytomy of three lineages, the B. distachyon 2n = 20 clade, B. mexicanum, and a strongly supported clade of the sister B. distachyon 2n = 10 + 2n = 30 and the core perennial clade. Within the last-named group there was a basal polytomy of B. arbuscula and B. retusum and a clade of more recently evolved taxa. The ETS topology (Fig. 6D, Supplementary Data Text) also favoured the early split of B. boissieri, followed by the successive divergence of B. distachyon 2n = 20 and B. mexicanum, although these were poorly supported, followed by the strongly supported divergence of the B. distachyon 2n = 10 + 2n = 30 and core perennial lineages. Within the latter B. distachyon group, all the identical 2n = 10 sequences and the 2n = 30 Bdistachyon402 sequence joined in a subclade The core perennial clade showed the successive strongly supported divergence of B. arbuscula and B. retusum, and a polytomy for the recently evolved taxa. The combined analyses of concatenated ITS and ETS sequences recovered a consensus nuclear rDNA topology (results not shown) that was mostly coincident with that of the ITS tree, supporting the early split of B. boissieri and then the unresolved divergence of B. distachyon 2n = 20, B. mexicanum and the B. distachyon 2n = 10 + 2n = 30/core perennials clade, indicating that the diploid B. distachyon 2n = 10 and its derived 2n = 30 allotetraploid were more closely related to the recently evolved long-rhizomatous perennials and less related to the apparently more ancestral diploid B. distachyon 2n = 20.

The nuclear single-copy CAL, DGAT and GI phylogenies were overall congruent with both the plastid and the nuclear multicopy rDNA trees and provided more detailed insights into the evolutionary origins of the B. distachyon cytotypes. The three genes undisputedly showed that the different allelic copies detected within B. distachyon 2n = 30 had been inherited from two different ancestors as they were nested, respectively, within the B. distachyon 2n = 20 and B. distachyon 2n = 10 clades. The CAL tree (Fig. 6E, Supplementary Data Text) supported an early split of B. boissieri, followed by that of the sister B. mexicanum (A)/B. retusum (2A) and then by a polytomy of sub-basal B. mexicanum (B), B. distachyon 2n = 20 (Bdistachyon114)/B. distachyon 2n = 30 (Bdistachyon113A) and the remaining taxa. The sister relationship of the B. distachyon 2n = 10/B. distachyon 2n = 30 (Bdistachyon113B) clade to the recently evolved core perennial clade was weakly supported. Within the last clade there was a further polytomy of three subclades that recovered various relationships among the different allelic copies of the perennial taxa. The DGAT tree (Fig. 6F, Supplementary Data Text) was unresolved for the basal radiation of the B. mexicanum, B. distachyon 2n = 20 (Bdistachyon114)/B. distachyon 2n = 30 (Bdistachyon113A) and further lineages. The sister relationship of the B. boissieri clade to B. retusum (A) was strongly supported, but that of this clade to the strongly supported B. distachyon 2n = 10 (Bdistachyon21)/B. distachyon 2n = 30 (Bdistachyon113D) clade had low support. The core perennial clade showed a divergence order that mostly agreed with that of the rDNA tree, although different allelic copies of B. retusum and B. pinnatum 2n = 28 fell within different subclades. The GI topology (Fig. 6G, Supplementary Data Text) showed the consecutive weakly supported splits of the B. boissieri (A, B, C) + B. retusum (A) clade and B. mexicanum (A), followed by the better supported ones of the sister B. mexicanum (B)/B. distachyon 2n = 20 (Bdistachyon114) + B. distachyon 2n = 30 (Bdistachyon113A), and the more recent split of the B. distachyon 2n = 10 (Bdistachyon21, Bdistachyon1) + B. distachyon 2n = 30 (Bdistachyon113B) clade and the largely unresolved core perennial clade.

Using a total-evidence principle approach, the combined plastid and nuclear rDNA tree (Fig. 6H) recovered a species-tree-like topology highly congruent with the nuclear data, based on the higher number of rDNA-informative characters over the plastid ones, in which the sister relationship of B. distachyon 2n = 10 (and its derived B. distachyon 2n = 30) clade to the core perennials clade was relatively well supported and there was a lack of resolution for the B. distachyon 2n = 20 to its closer relatives.

Our divergence time estimates (Beast mean and interval values; Fig. 6H) suggested that the Brachypodieae and the core pooid clade diverged from their MRCA at (25·7) 29·8 (33·6) Ma, and that the successive splits of B. boissieri, B. distachyon 2n = 20/B. mexicanum/remaining lineages, B. distachyon 2n = 10 + 2n = 30/core perennials, B. distachyon 2n = 10/(derived) B. distachyon 2n = 30, core perennials, and B. distachyon 2n = 30 lineages took place at (7·6) 13·3 (21·0) Ma, (5·8) 9·9 (16·4) Ma, (3·8) 7·2 (12·0) Ma, (2·0) 4·3 (7·8) Ma, (1·1) 2·9 (6·2) Ma and (0·3) 1·0 (2·6) Ma, respectively. Divergence times for the B. distachyon 2n = 20 and B. distachyon 2n = 10 crown groups could not be estimated with confidence due to their respective low intragroup variation found within the present sampling.

Assuming that the divergence of the different Brachypodium lineages occurred at the estimated dates, the average calibrated substitution rates of the main groups for the nuclear rDNA and plastid genes ranged from the highest rates of the basal B. boissieri and B. mexicanum lineages [rDNA: 3·75 × 10⁻⁹ to 1·36 × 10⁻⁹ substitutions per site per year (s/s/y), with a mean of 2·14 × 10⁻⁹ s/s/y; cpDNA: 3·75 × 10⁻⁹ to 1·36 × 10⁻⁹ (mean 2·14 × 10⁻⁹) s/s/y] to the lowest rates of the recently evolved core perennial lineage [rDNA: 2·73 × 10⁻⁹ to 4·84 × 10⁻¹⁰ (mean 1·03 × 10⁻⁹) s/s/y; cpDNA: 2·73 × 10⁻⁹ to 4·84 × 10⁻¹⁰ (mean 1·03 × 10⁻⁹) s/s/y], with average rate values of 1·90 × 10⁻⁹ to 6·90 × 10⁻⁹ (mean 1·09 × 10⁻⁹) s/s/y (rDNA) and 9·87 × 10⁻¹⁰ to 3·57 × 10⁻¹⁰ (mean 5·64 × 10⁻¹⁰) s/s/y (cpDNA) for the whole Brachypodium clade (Table 3A). Comparisons of the number of nucleotide substitutions between the most divergent related pairs of species of each group and their consensus divergence from the relevant MRCA indicated the occurrence of different substitution rates, with the highest rDNA rates found in the comparisons of B. distachyon 2n = 20 – B. boissieri (mean 3·16 × 10⁻⁹ s/s/y) and B. distachyon 2n = 20 – B. distachyon 2n = 10 (mean 2·98 × 10⁻⁹ s/s/y) pairs, and the highest cpDNA rate in the B. boissieri – B. mexicanum (mean 1·73 × 10⁻⁹ s/s/y) pair (Table 3A).

Table 3.

(A) Absolute substitution rates (substitutions per site per year) of the main Brachypodium groups obtained from averaged calibrated values within each group and between distant pairs of species of different groups for the nuclear ribosomal (rDNA) and plastid (cpDNA) genes

Groups – Species pairs	rDNA			cpDNA
	Min.	Mean	Max.	Min.	Mean	Max.
Groups:
Brachypodium	1·90 × 10−9	1·09 × 10−9	6·90 × 10−10	9·87 × 10−10	5·64 × 10−10	3·57 × 10−10
Basal perennials	3·75 × 10−9	2·14 × 10−9	1·36 × 10−9	3·75 × 10−9	2·14 × 10−9	1·36 × 10−9
B. distachyon20 + basal perennials	3·16 × 10−9	1·80 × 10−9	1·14 × 10−9	9·21 × 10−10	5·26 × 10−10	3·33 × 10−10
B. distachyon(10 + 30) + core perennials	2·90 × 10−9	1·53 × 10−9	9·17 × 10−10	1·58 × 10−9	8·33 × 10−10	1·00 × 10−10
B. distachyon(10 + 30)	1·98 × 10−10	1·53 × 10−10	9·17 × 10−10	1·90 × 10−9	1·11 × 10−9	6·71 × 10−10
Core perennials	2·73 × 10−9	1·03 × 10−9	4·84 × 10−10	2·73 × 10−9	1·03 × 10−9	4·84 × 10−10
Distant species pairs:
B. distachyon20 – B. distachyon10	5·09 × 10−9	2·98 × 10−9	1·80 × 10−9	2·15 × 10−9	1·26 × 10−9	7·62 × 10−10
B. distachyon20 – B. boissieri	5·52 × 10−9	3·16 × 10−9	2·00 × 10−9	1·05 × 10−9	6·01 × 10−10	3·81 × 10−10
B. distachyon20 – B. mexicanum	3·15 × 10−9	1·80 × 10−9	1·14 × 10−9	1·12 × 10−9	6·56 × 10−10	3·96 × 10−10
B. distachyon20 – B. sylvaticum	4·74 × 10−9	2·78 × 10−9	1·68 × 10−9	1·72 × 10−9	1·01 × 10−9	6·10 × 10−10
B. distachyon10 – B. boissieri	4·14 × 10−9	2·37 × 10−9	1·00 × 10−9	1·12 × 10−9	6·39 × 10−10	4·05 × 10−10
B. distachyon10 – B. mexicanum	2·15 × 10−9	1·26 × 10−9	7·62 × 10−10	1·46 × 10−9	8·58 × 10−10	5·18 × 10−10
B. distachyon10 – B. sylvaticum	3·55 × 10−9	1·87 × 10−9	1·12 × 10−9	3·03 × 10−9	1·60 × 10−9	9·58 × 10−10
B. boissieri – B. mexicanum	3·16 × 10−9	1·80 × 10−9	1·14 × 10−9	3·03 × 10−9	1·73 × 10−9	1·09 × 10−9
B. boissieri – B. sylvaticum	3·88 × 10−9	2·22 × 10−9	1·40 × 10−9	9·21 × 10−10	5·26 × 10−10	3·33 × 10−10
B. arbuscula – B. sylvaticum	1·82 × 10−9	6·90 × 10−10	3·22 × 10−10	5·00 × 10−9	1·90 × 10−9	8·87 × 10−10

Absolute rates were estimated in each case from averaged K_s and minimum (min.), mean and maximum (max.) divergence times from their corresponding MRCAs. Basal perennials: B. boissieri, B. mexicanum. Core perennials: B. arbuscula, B. phoenicoides, B. pinnatum, B. retusum, B. rupestre, B. sylvaticum.

(B) Relative-rate likelihood ratio test values of pairwise comparisons of Brachypodium spp. nuclear ribosomal DNA (lower left matrix) and plastid DNA (upper right matrix) sequences using Secale cereale as outgroup

rDNA	cpDNA
	Bsyl	Brup	Bret	Bpin	Bpho	Bmex	Bdis20	Bdis10	Bboiss	Barb
B. sylvaticum	–	0·89	3·55+	0·2	<–0·01	0·69	2·01	0·74	0·25	0·02
B. rupestre	2·58	–	1·27	1·22	2·45	0·05	1·21	1·3	0·02	0·74
B. retusum	1·86	0·51	–	2·42	3·62+	<0·01	2·77	1·53	0·06	1·99
B. pinnatum	0·28	0·47	0·96	–	0·34	0·36	2·69	1·15	0·07	0·26
B. phoenicoides	<–0·01	2·57	1·87	0·28	–	0·69	2·01	0·73	0·25	0·02
B. mexicanum	7·30**	5·62*	4·29*	6·81**	7·31**	–	4·50*	2·01	<0·01	0·64
B. distachyon20	3·39+	4·67*	5·65*	3·79+	3·39+	16·14***	–	0·24	2·47	0·67
B. distachyon10	0·16	<0·01	0·11	0·05	0·16	4·51*	4·31*	–	1·25	0·3
B. boissieri	1·71	1·31	0·89	1·38	1·71	0·12	8·11**	1·3	–	0·31
B. arbuscula	<–0·01	0·17	0·59	0·09	<0·01	5·67*	3·82+	0·01	1·68	–

The ratio is given as the ratio of row over column (rDNA) and of column over row (cpDNA). Significant (*P < 0·05; **P < 0·01; ***P < 0·001) or marginally significant (⁺P = 0·5–0·6) values are indicated in bold.

The results obtained from the RRTs are shown in Table 3B. The more variable nuclear ribosomal data set detected significant differences in the evolutionary rates of B. mexicanum and B. distachyon 2n = 20 with respect to those of most of the other Brachypodium taxa whereas the more conserved substitutional rate of the plastid region did not detect significant differences among them except for that between B. distachyon 2n = 20 and B. mexicanum and marginal alterations between B. retusum – B. sylvaticum and B. phoenicoides – B. retusum. The short-rhizomatous B. mexicanum showed accelerated rDNA mutation rates, significantly different from all others except B. boissieri. Similarly, the annual B. distachyon 2n = 20 showed faster mutation rates, significantly (or marginally significantly) different from all the perennial species and also from that of the annual B. distachyon 2n = 10. Interestingly, the latter species did not show significant differences in evolutionary rate with respect to those of the perennial Brachypodium taxa (Table 3B).

DISCUSSION

Taxonomic split of the Brachypodium distachyon complex: B. distachyon, B. stacei, B. hybridum

The multidisciplinary approach undertaken in this study has clearly demonstrated that the three cytotypes of B. distachyon do not correspond to an autopolyploid series falling within the range of variability of a single species but, rather, to three well-differentiated taxa, each of them deserving the rank of species. The cytogenetic and molecular data concur in that the two diploid B. distachyon 2n = 10 and 2n = 20 cytotypes represent divergent species that participated in the origin of the derived allotetraploid 2n = 30 cytotype, whereas the morphoanatomical data have shown phenetic differences between the two diploids, with the derived allotetraploid having acquired some of its own phenetic attributes while sharing others with one or other parental lineage.

The cytogenetic analyses unambiguously support the distinct genomic nature of the diploid B. distachyon 2n = 10 and 2n = 20 taxa, containing chromosomes that do not cross-hybridize in GISH experiments, and showing different base numbers of chromosomes (x = 5 and x = 10) with distinctly different morphometric features. Furthermore, this is strongly supported by FISH mapping of 5S and 25S rDNA as well as numerous ‘single-locus’ clones from the B. distachyon (2n = 10) BAC-library (Fig. 4; Hasterok et al., 2004, 2006). This clearly shows that the B. distachyon 2n = 30 taxon is a derived allotetraploid of the 2n = 10 and 2n = 20 taxa, as demonstrated by its cross-hybridizing GISH and FISH signals in its distinctly inherited 10- and 20-chromosome sets (Fig. 4), their correlated chromosome sizes, and its genome size of 1·265 pg/2C (this study), resulting from the approximate sum of the genome sizes of the two diploid 2n = 10 (0·631 pg/2C DNA, Wolny and Hasterok, 2009) and 2n = 20 (0·564 pg/2C, this study). Also, the results of CCP clearly show that the painted bivalents seen in B. distachyon 2n = 30 represent the simple sum of painted bivalents observed in B. distachyon 2n = 10 and 2n = 20 (Fig. 5; Idziak et al., 2011). Our cytogenetic approach concurs with other grass studies that have demonstrated the importance of using in-situ hybridization to accurately identify the diploid progenitors of allopolyploids (e.g. Zingeria, Kotseruba et al., 2010).

The evolutionary study has further demonstrated that previously unrecognized diagnostic morphological differences within the annual B. distachyon s.l. taxa correspond to two highly divergent ancestral diploid lines and their derived allotetraploid, which show greater divergence and higher substitution rates than the morphologically well-characterized but less divergent and recently evolved core of perennial Brachypodium taxa (Table 3, Fig. 6). The phylogenetic analyses based on maternally inherited plastid genes have further confirmed that the ancestral B. distachyon 2n = 20 taxon was the maternal donor of the derived allotetraploid whereas the phylogenetic analyses based on the biparentally inherited nuclear genes have confirmed that the apparently less ancestral B. distachyon 2n = 10 taxon was its paternal donor. Analysis of the nuclear ribosomal ITS and ETS regions, which may be prone to convergent evolution, has shown that the ribosomal sequences of the allotetraploid B. distachyon 2n = 30 taxon resemble those of its B. distachyon 2n = 10 paternal parent, whereas analysis of the low-copy CAL, DGAT and GI genes indicates that the allotetraploid maintains the orthologous loci of both parents. This agrees with previous studies of other temperate grasses (e.g. Elymus) in which the ribosomal ITS sequences of the derived allopolyploids have converged into the sequences of one parental lineage whereas the two distinct parental copies of single-copy genes are still present in the amphidiploid hybrids (Mahelka and Kopecky, 2010; Mason-Gamer et al., 2010).

The large phenotypic variability of the B. distachyon s.l. complex taxa has not passed unnoticed, and both taxonomic entities (cf. Lahondère, 1985; Schippmann, 1991) and cytogenetic or plant-breeding races (Garvin et al., 2008; Vogel et al., 2009) have been recognized in the literature. However, no comprehensive taxonomic treatment has been proposed before. Our systematic and evolutionary studies have demonstrated that the B. distachyon complex is composed of three different species, showing a clear speciation pattern of two relict diploids and a more recently derived allotetraploid. Because the B. distachyon 2n = 10 taxon has been selected as model grass species for the temperate cereals (IBI, 2010; Mur et al., 2011, and references therein) it is appropriate to describe the other 2n = 20 and 2n = 30 taxa as representing novel species, as they form a hybridizing diploid-to-polyploid evolving complex that parallels the origins of the most economically important Triticeae, Loliinae and Aveneae species and could therefore serve as an ideal model for wheat, oats and temperate forage grasses.

The completion of the Brachypodium whole-genome sequencing project has spawned considerable scientific interest, with many world-class laboratories currently working on the genomics, transcriptomics, proteomics and metabolomics of the sequenced 2n = 10 B. distachyon line Bd21 (http://www.brachypodium.org/). Due to its current wide usage, it is imperative to keep the specific name distachyon for the diploid 2n = 10 taxon. However, the intermediate phenotypic features of the B. distachyon LINN 93·48 type specimen between the 2n = 20 and 2n = 10 taxa (Fig. 2B) and the impossibility to analyse this specimen cytogenetically and evolutionarily precludes its use as the type specimen of this plant. Hence, we chose the diploid B. distachyon 2n = 10 line Bd21 specimen as an epitype (Art. 9·7 ICBN; McNeill et al., 2005). This will maintain the naming of the most common and widespread of the two possible taxa, which is today an important model plant for studies of the genomics of cereal relatives. We have also selected holotype specimens for the newly described 2n = 20 and 2n = 30 species from the respective inbred line ABR114 and ABR113 materials, that show distinctive genetically fixed phenotypic features which have been biologically validated by the study of wild individuals (Table 2, Fig. 2A, B).

The completely sequenced 2n = 10 B. distachyon genome has provided an extraordinary amount of information for current comparative phylogenomic analyses of this plant and other totally or partially sequenced grasses. In the same way, the availability of fully sequenced B. distachyon-like 2n = 20 B. stacei and 2n = 30 B. hybridum would provide an exceptional opportunity for thorough next-generation phylogenomic analysis of genetically divergent taxa.

Taxonomy

Brachypodium distachyon (L.) P. Beauv., Essai Agrost. 101, 155, 156. 1812. Bromus distachyos L., Fl. Palaest. 13. 1756.

 Festuca distachya (L.) Roth, Catal. Bot. 1: 11. 1797.

 Agropyron distachyon (L.) Chevall., Fl. Gén. Env. Paris 2: 196. 1827.

 Trachynia distachya (L.) Link, Hort. Reg. Bot. Berol. 1: 43. 1827.

 Zerna distachya (L.) Panz. ex B.D. Jacks., Index Kew. 2: 1249. 1895.

TYPE: LINN 93·48 (neotype designated by Schippmann & Jarvis, Taxon 37: 158, f. 1 (1988). EPITYPE (designated here): Iraq: Salah ad Din: 4 km from Salahuddin, in the road to Mosul. Col. No. K1202. USDA PI 254867, Bd21 inbred line, from seeds cultivated at Aberystwyth University, 30 October 2010, collector Luis Mur (epitype MA, isoepitypes JE, K).

Flowering culms (18)26(35) cm tall, glabrous or nearly so, with (2)3(5) nodes, nodes hairy. Culm leaf blades flat, (4·3)7(8·5) cm long, 0·25–0·33 cm wide. Leaf-sheaths glabrous or pubescent. Leaf blades usually with scattered short and long hairs on the adaxial and abaxial sides, the short hairs sometimes absent. Stomata guard cells (18)23(29) μm long. Panicles (1·5)3(4·8) cm long, rachis glabrous, with (1)3(4) spikelets. Spikelets (12·5)16(22) mm long, (12)13(15) mm to the apex of the fourth lemma, with (5)7(9) florets. Glumes oblong-lanceolate, shorter (more than half) than the contiguous florets, abaxially glabrous to densely long hairy, the distal (and longer) glumes (6·2)7·2(8·2) mm long. Lemmas (7)8(9) mm long, lanceolate, abaxially nearly glabrous to densely long hairy; awn (9·5)11·5(13) mm long, as long as or longer than lemma, hairy. Anthers 2–3, 0·5–0·7 mm long, non-exerted. Pollen grains (27)30(35) μm long. Ovaries oblong, hairy at apex. Caryopsis (4·8)6·2(7·2) mm long, oblong. Flowering time (months): IV–VI (VII).

Native to the Mediterranean region (France, Iraq, Italy, Slovenia, Spain, Turkey, probably spread in southern Europe, south-west Asia and North Africa). Common in mountain areas, between 300 and approx. 1700 m altitude, growing in open woods and ephemeral pastures.

Diploid with 2n = 10 chromosomes (x = 5). Genome size of 0·631 pg/2C. Relatively large chromosomes.

Brachypodium stacei Catalán, Joch. Müll., Mur & Langdon sp. nov.

TYPE: Spain: Balearic Isles: Formentera: Torrent, ABR114 inbred line, from seeds cultivated at Aberystwyth University, 30 October 2010, collector Luis Mur (holotype MA, isotypes JE, K). It differs from Brachypodium distachyon in its chromosome number of 2n = 20 and its often longer spikelets and anthers. [Differt a Brachypodio distachyo chromosomatum numero 2n = 20 et spiculis antherisque saepe longioribus.]

Flowering culms (26)36(46) cm tall, pubescent in the basal half, with (2)3(4) nodes, nodes hairy. Culm leaf blades flat, 7–14 cm long, 0·26–0·7 cm wide. Leaf-sheaths pubescent. Leaf blades usually with scattered short and long hairs. Stomata guard cells (20)28(38) μm long. Panicles (2·1)6·2(9·5) cm long, rachis glabrous, with (1)3(4) spikelets. Spikelets (15)22(41) mm long, (13)15(30) mm to the apex of the fourth lemma, with (6)9(14) florets. Glumes oblong-lanceolate, shorter (more than half) than the contiguous florets, abaxially glabrous or nearly so, although often with ciliate margins, the distal glumes (5)6(7·3) mm long. Lemmas (7·4)9(11·2) mm long, lanceolate, abaxially glabrous or nearly so, but often with ciliate margins, awn (2·3)7(10·2) mm long, as long as or longer than lemma, hairy. Anthers 3, 0·6–0·8 mm long, non-exserted. Pollen grains (26)33(38) μm long. Ovary oblong, hairy at apex. Caryopsis (6·1)6·2(6·3) mm long, oblong. Flowering time (months): IV–VI (VII).

Native to the western Mediterranean region (France, Portugal, Spain). Rare. Mostly found in islands (Balearic Isles, Corsica). Also cited from Sardinia, Sicily, southern Spain, southern Italy and Morocco (Lahondère, 1985; Hammami et al., 2011) and western China (Shang et al., 2011), between sea level and unknown upper altitudinal limit, growing in sand beaches and ephemeral pastures.

Diploid with 2n = 20 chromosomes (x = 10). Genome size of 0·564 pg/2C. Small chromosomes. Species dedicated to Prof. Clive A. Stace, who initiated the systematic and evolutionary studies of Brachypodium.

Brachypodium hybridum Catalán, Joch. Müll., Hasterok & Jenkins sp. nov.

TYPE: Portugal: Lisbon, ABR113 inbred line, from seeds cultivated at Aberystwyth University, 30 May 2011, collector Tim Langdon (holotype MA, isotypes JE, K). It differs from Brachypodium distachyon in its chromosome number of 2n = 30 and its longer lemmas and often longer spikelets and distal glumes. [Differt a Brachypodio distachyo chromosomatum numero 2n = 30 et lemmis et saepe glumis distalibus spiculisque longioribus.]

Flowering culms (17)42(48) cm tall, pubescent in the basal half or for their full lengths, with (1)4(7) nodes. Culm leaf blades flat, (6)11(16) cm long, (0·11)0·31(0·4) cm wide. Leaf-sheaths pubescent. Leaf blades usually with scattered short and long hairs. Stomata guard cells (28)34(40) μm long. Panicles (2)4·4(5·5) cm long, rachis glabrous to densely long hairy, with (1)3(5) spikelets. Spikelets (15)22(38) mm long, (13·8)16(18) mm to the apex of the fourth lemma, with (5)8(16) florets. Glumes oblong-lanceolate, shorter (more than half) than the contiguous florets, abaxially glabrous to densely long hairy, the distal glumes (7)8(9·7) mm long. Lemmas (9)10·5(12·4) mm long, lanceolate, green, abaxially glabrous to densely long hairy; awn (8)10·5(12·2) mm long, as long as or longer than lemma, hairy. Anthers 3, 0·5–1 mm long, non-exserted. Pollen grains (34)38(48) μm long. Ovaries oblong, hairy at apex. Caryopsis (5·7)7(8·2) mm long, oblong. Flowering time (months): IV–VI (VII).

Native to the Mediterranean region (Afghanistan, Iran, France, Morocco, Pakistan, Spain, probably spread in southern Europe, south-west Asia and North Africa). Introduced in central and western Europe, Australia, North and South America, and South Africa. Common in lowland and coastal areas, between 0 and 2000 (3600) m altitude, growing in open and disturbed areas.

Allotetraploid with 2n = 30 chromosomes (x = 10 + x = 5), derived from the cross between B. stacei and B. distachyon. Genome size of 1·265 pg/2C. Contains both relatively small B. stacei-type and large B. distachyon-type chromosomes. The name hybridum refers to its hybrid allotetraploid nature.

The taxonomic characterization of the three species is based on the informative phenetic characters that have shown significant differences between these taxa and that could be used as diagnostic traits (Tables 2 and 3, Fig. 1). The most distinct B. distachyon could be separated from B. stacei and B. hybridum based on its often smaller height (<32 cm), shorter culm leaves (<9 cm), shorter spikelets (<20 mm, total; <14 mm up to the 4th floret) and shorter lemma (<8·7 mm), whereas B. stacei could be separated from B. hybridum based on its shorter upper glume (<7 mm), shorter lemma (<9·8 mm), shorter awn (<10 mm) and shorter caryopsis (<6·3 cm). The microtaxonomic characters related to leaf-stomata guard cell length and pollen length that have shown to be useful to discriminate among the three B. distachyon complex species have also proved to be informative to separate other close diploid and polyploid grasses (Borrino and Powell, 1988; Katsiotis and Forsberg, 1995). Interspecific gene flow does not apparently occur (or might be extremely low) among the three taxa, even for those taxa that may be sympatric (e.g. B. distachyon and B. hybridum), due to the strong inbred nature of B. distachyon. Also, synthetic B. hybridum-like hybrids are difficult to produce as a consequence of this feature of B. distachyon (G. Linc, Hungarian Academy of Sciences, Agricultural Research Institute, Martonvasar, Hungary, pers. comm.), adding further support to the species ranking proposed for them.

Evolution and speciation of the annuals B. stacei, B. distachyon and B. hybridum

Although the differentiation of the B. stacei, B. distachyon and B. hybridum species has been positively confirmed by all phenotypic, cytogenetic and molecular approaches, the evolutionary history of these annual Brachypodium taxa has not been completely resolved. All the plastid and nuclear phylogenetic trees agree in the ancestral divergence of B. stacei and in the relatively ancestral one of B. distachyon (Fig. 6). However, most plastid and nuclear genes show different relationships of these annual lineages to distinct perennial lineages (Fig. 6A–G). The plastid data (mostly ndhF) recovered a sister relationship of B. stacei to B. mexicanum and of B. distachyon to B. boissieri (and then to the core perennials clade), whereas the rDNA, CAL and GI data placed B. stacei in a sub-basal unresolved position and B. distachyon strongly to weakly related to the recently evolved perennials. By contrast, the DGAT data recovered weakly supported relationships of B. distachyon to a sister ancestral B. boissieri/B. retusum p.p. clade (Fig. 6F). These apparently incongruent topological nestings might be the consequence of a rapid radiation of the basal Brachypodium lineages, which resulted in the observed incomplete lineage sortings among the studied genes, or of ancestral hybridization processes between the annuals and the perennials. The overall poorly supported splits of the basal and sub-basal Brachypodium lineages could explain the ambiguous resolutions obtained for B. stacei, B. mexicanum, B. boissieri and B. distachyon. Nonetheless, five of the seven analysed genes support the sister relationship of B. distachyon to the recently evolved clade of core perennials (Fig. 6A, B, D–E, G). The trees also agree in that some perennial lineages, such as B. mexicanum and B. boissieri in the sampled plastid and rDNA trees, and these taxa plus one copy of B. retusum in the non-concerted single-copy CAL, DGAT and GI trees, also evolved from the most basal Brachypodium ancestors (Fig. 6). Wolny et al. (2011) combined cytogenetic and CAL, GI and STT3-based phylogenetic studies to suggest the existence of an ‘ancestral’ diploid Brachypodium genome that would be present in B. mexicanum, B. retusum p.p. and B. distachyon s.l. p.p., and a more recently evolved one that gave rise to B. distachyon p.p. and the core perennial species. These authors also suggested a homoploid hybrid origin for the 2n = 10 diploid B. distachyon from the ancestral and recent genomes, a result that was supported by a mosaic pattern of alternate gene phylogenies across a relatively large chromosomal contig. Our current results add support to this hypothesis and further confirm that the ribosomal DNA sequences of B. distachyon might have evolved in a concerted way towards those of the recent perennial genome rather than to the ancestral genome ones. This is corroborated by the lack of significant differences in evolutionary rDNA rates between B. distachyon and the core perennial group (Table 3B) and by the lower absolute rDNA rates observed in B. distachyon and the core perennials (mean 1·53 × 10⁻⁹ to 1·03 × 10⁻⁹ s/s/y) in contrast to the higher values of the basal and sub-basal B. boissieri, B. mexicanum and B. stacei lineages (mean 2·14 × 10⁻⁹ to 1·80 × 10⁻⁹ s/s/y) (Table 3A).

The potential basal divergence of an x = 5 lineage suggested by some early nuclear marker topologies (Catalán et al., 1995) has not been confirmed by the sequence data as none of the recovered trees placed B. distachyon sister to B. mexicanum. By contrast, some topologies support a strong sister relationship of B. mexicanum to B. stacei (e.g. ndhF, Fig. 6A), suggesting that the reduction of the chromosome base number from x = 10 to x = 5 could have occurred at least twice or that the chromosome base number of B. mexicanum is not x = 5 but x = 10. This might be indirectly corroborated by the absence of multiple allelic copies of the single-copy genes CAL, DGAT and GI in B. mexicanum, denoting an unlikely pattern for an octoploid. The results of CCP (Fig. 5) support our previous hypothesis (Hasterok et al., 2006) that multiple centric fusion/fission events could be one of the mechanisms responsible for the present structure of B. stacei and B. distachyon karyotypes. In the light of results of the recent cytomolecular studies (Idziak et al., 2011; Wolny et al., 2011) it is possible, however, that individual chromosomes in the karyotypes have evolved in different ways, involving various small-scale chromosome rearrangements, such as duplications or translocations. The successive divergence of B. arbuscula, B. retusum and other perennials within the core perennial clade found in almost all the analysed trees agree with those of Catalán and Olmstead (2000) and current investigations of the whole genus (P. Catalán et al., unpub. res.), suggesting that several reductions from x = 10 to x = 9, 8 and 7 and multiple polyploidization events within the x = 8 and 7 taxa might have also occurred in the evolutionary history of these more recent perennial lineages. Evolutionary rate tests indicated that the accelerated mutation rates of the rDNA sequences of B. mexicanum and B. stacei are significantly faster than almost all the others and also significantly different from each other. However, those of the most basal B. boissieri are not (Table 3B). This suggests distinct adaptive scenarios for the evolution of the fast mutating short-rhizomatous American B. mexicanum and annual Mediterranean B. stacei, which could have colonized more rapidly open xeric grassland habitats, with respect to the slow mutating long-rhizomatous strict endemic B. boissieri, which would have settled down in more stable shrubby habitats. The latter evolutionary scenario parallels that of the also slow mutating remaining long-rhizomatous perennial species that have colonized other stable mesic grassland, shrubby and woodland habitats.

Although the two annual diploid B. stacei and B. distachyon lineages did not apparently derive from the same MRCA, the origins of these two species had to precede that of their derived allotetraploid B. hybridum. The dating analysis indicates that, despite the late Oligocene age of the stem Brachypodieae line (approx. 29·8 Ma), the radiation of the crown node occurred in middle Miocene times (approx. 13·3 Ma). The dated divergence indicated that B. stacei and other ancestral lineages probably originated in the middle Miocene (Tortonian, approx. 9·9 Ma) whereas B. distachyon and the core perennials split from their MRCA in the late Miocene (Messinian, approx. 7·2 Ma). Although the ages of radiation of B. stacei and B. distachyon could not be confidently estimated, the dating analysis suggested that B. hybridum could have arisen from the cross between them in more recent Pleistocene times (approx. 1 Ma). The extraordinarily accelerated mutation rates observed in the nuclear rDNA sequences of the annual B. stacei, and to lesser extent in B. distachyon and their derived B. hybridum, and the significant differences of the former with respect to the decelerated mutation rates of the perennials (Table 3) are the likely result of their shorter life cycles but also of a long history of isolation. As observed in other grasses (Gaut et al., 1996, 1997; Catalán et al., 2006), the minimum generation time is the main factor determining the evolutionary rates of the fast-evolving annuals and the slow-evolving perennials. In this case, B. stacei showed greater mutation rates than those of its close perennial B. boissieri. An independent acquisition of the annual life cycle in the B. stacei and B. distachyon lineages would have fostered their accelerated mutation rate relative to the typical average of the perennial species. This could be of great importance as the higher mutation rates might have conferred a higher environmental adaptive capability to the annual species, which could be confirmed in the future through comparative phylogenomic analyses of the three annual species and of some perennials. Additionally, another noticeable finding of our molecular studies was the apparent low infraspecific variability of B. stacei, B. distachyon and B. hybridum. Despite the restricted number of samples included in this study, individuals from geographically separated populations of each species presented identical or slightly divergent sequences (see Supplementary Data Text, available online). This might be related to their inbred nature, which enables the generation of almost pure homozygous populations in a few generations (Vogel et al., 2009), together with an easy dispersal mechanism of seeds, which probably allowed their colonization of the Mediterranean region. Current investigations using highly variable markers and a broad geographical sampling are under way to establish the spatial genetic structure of the three species and to determine the potential single vs. multiple origins of the allotetraploid B. hybridum from its B. stacei and B. distachyon parents.

SUPPLEMENTARY DATA

Supplementary data are available online at http://www.aob.oxfordjournals.org and consist of the following. Table S1: GenBank accession numbers of the Brachypodium spp. and outgroup samples used in the phylogenetic analyses. Supplementary Text: features of the plastid (ndhF, trnLF) and nuclear (ITS, ETS, CAL, DGAT, GI) data sequences used in the evolutionary analysis and on the parameters of the optimal phylogenetic trees obtained from them.