Plant material

Chromosome and probe preparation

All seeds were germinated on moistened filter paper at 27°C in Petri dishes until the roots were 2 cm long, and then treated with 2 mM 8-hydroxyquinoline for 2 h at room temperature to accumulate metaphases. After rinsing with distilled water, the whole seedlings were fixed in a mixture of freshly prepared 1:3 glacial acetic and ethanol (100%), and stored at −20°C until use. For DNA probes, plants were grown in autoclaved soil, transferred into an open air field at the seedling stage, and DNA extracted from young leaves for PCR amplification of 5S rDNA gene and knotted1 gene fragments and for probe preparation for GISH.

Chromosome preparations were done according to Benabdelmouna et al [30], with some modifications. A single root tip was transferred in a drop of 45% acetic acid onto a clean slide before squashing. After squashing, the slides were then scanned for good chromosome spreads at prophase or metaphase stages with phase contrast microscopy, then slips were removed by immersing slides into liquid nitrogen for 10–15 min, and the air-dried slides were stored at room temperature until use.

In situ hybridization

Phylogenetic analysis

Total genomic DNA was extracted from fresh leaves using the CTAB procedure [32]. The 5S rDNA gene fragments from each sample were amplified using primers designed on conserved regions of the 5S rDNA sequences from barley and wheat [33] with FP: 5^′-GGACCTCCTGCGAAGTCCT-3^′ and RP: 5^′-CCCATCCGTGTACTACTCTC-3^′. The PCR conditions were 95°C for 5 min, 32 cycles of 94°C for 55 s, 62°C for 25 s and 72°C for 35 s, followed by a final extension of 72°C for 10 min. The PCR mixture (20 μl) contained 100 ng template DNA, 250 μM of each dNTP (Takara), 0.5 μM of each primer, 2.0 μl of reaction buffer, and 1 unit of Ex Taq DNA polymerase (Takara). The knotted1 (kn1) gene fragments including part of the first intron, the whole sequences of the second intron and the second exon as described by Doust et al [9] were amplified using the primers pair of kn211-402F: 5^′-TCAGAACTTTTGGCCGTGGGT-3^′ and kn612-402R: 5^′-GAGATGGACAGCGAGTTGAGC-3^′. PCR reactions were denatured for 5 min at 95°C and then followed a step-down procedure where annealing temperature was stepped down from 65°C to 59°C, and then 35 amplification cycles were performed, each cycle including denaturation at 95°C for 30 s, annealing at 57°C for 1 min, primer extension at 72°C for 1 min, and a final extension of 72°C for 10 min. Products were separated by electrophoresis on 1.5% agrose gels, and the major band of each sample was isolated from the gel, cleaned using Qiagen columns, cloned into the PUCm-T or PMD18-T vector system, and then transformed into the DH5a strain of E. coli. After identification of recombinant clones, for diploid species, a minimum of two clones were sequenced, for tetraploid species, at least six clones were picked and sequenced. Kn1 and 5S rDNA fragments obtained were confirmed by comparison with Genbank. Sequences from the same accession were aligned using DNAMAN, and redundant sequences were deleted. Accession number for each clone has been deposited in Gene Bank and listed in Additional file 1: Table S1. Outgroup sequences [AB023618, DQ351339, JQ947589, X61308] were obtained from Genbank.

Multiple alignments of unique sequences from each sample were carried out using T-Coffee [34–36]. The phylogenetic trees based on 5S rDNA and kn1 sequences were analyzed under neighbor joining (Mega version 5.1) [37] and Bayesian approaches (Mr Bayes 3.1.2) [38, 39]. Evolutionary models were chosen using jModeltest version 0.1.1 [40, 41], and in both cases were a general time reversible model with a 4 category gamma rate model. Neighbor-joining analyses were run under the Maximum-composite likelihood model, with gamma values estimated in jModeltest, and tree support estimated with 1,000 bootstrapped sample sets. Mr Bayes was run using a GTR plus gamma model, with branch length unconstrained, branch length priors set to an exponential distribution with a parameter of 10, and shape parameter priors set to an exponential distribution with a parameter of 10. Bayesian analyses were run for 5,000,000 generations, with 4 chains on each of two nodes. Chains were compared every generation and nodes compared every 1,000 generations. Tracer version 1.5 [42] was used to estimate burnin, and a conservative 1,000,000 generations from the beginning of each run were removed. The remaining trees from each run were combined and a majority rule consensus tree used computed to summarize the data.

GISH analysis

Genome identification of a novel diploid Setariaspecies

GISH hybridization patterns were investigated using genomic DNA probes to mitotic chromosomes. When the probe prepared from total genomic DNA of S. virdis-Q24 was hybridized on the chromosome of S. italica, a strong and total painting on all the chromosomes of S. italica was observed (Figure 1a). However, when the S. viridis-Q24 genomic DNA was used as a probe on chromosome preparations of Qing 9, little hybridization was observed except for two major signal points in the nucleolar organizer region (Figure 1b). When total DNA probe from Qing 9 was applied to chromosome preparations of S. adhaerans-W94, all 18 chromosomes strongly hybridized (Figure 1c-d). The same hybridization pattern was obtained in the reciprocal experiment when genomic DNA from S. adhaerans-W94 was hybridized to chromosome preparations of Qing 9 (Figure 1e-f).

Phylogenetic analysis

Phylogeny based on 5SrDNAsequences

All PCR amplifications for 5S rDNA sequences from primers designed by D’Hont et al [33] were successful using our Setaria genomic DNA as template, and target bands were isolated and subcloned for sequencing. For diploid species, such as S. viridis, S. italica, S. adhaerans and S. grisebachii, single sequences were obtained from a sample accession, but for polyploid species, such as S. verticillata, S. faberi, S. glauca, S. parviflora, S. palmifolia, S. plicata and S. arenaria, multiple isolates were sequenced and sequences obtained were first aligned to delete redundant sequences, before the unique ones were used for phylogenetic tree construction. The number of unique sequences obtained varied between the different polyploid species (Table 1).

A consensus tree generated from the Bayesian analysis of the 5S rDNA sequences is displayed in Figure 2, and six groups can be clearly identified (similar results were found for the neighbor joining tree, although with less resolution). The first group, which corresponds with the A genome, is composed of sequences from S. viridis, S. italica, S. verticillata, S. faberi and S. queenslandica. The sequence of S. faberi-W7-37 was somewhat divergent from the others of the A group sequences and also carried a large deletion of 41 bp compared to the others in this group. Group B is composed of sequences from S. adhaerans, S. verticillata, and Qing 9, which corresponds with the B genome. Two sequences of S. verticillata deposited in Genbank [AF227011, AF227012] were separately included in A and B group. Group C is composed of two sequences, one from S. grisebachii and the other from S. lachnea, which corresponds with the known C genome. Although S. lachnea is a tetraploid, only one 5S rDNA was amplified. Sequences in group D were all from two tetraploid species of S. glauca and S. parviflora with high support, indicating a new genome that is different from the known A, B and C genomes. High support values were also seen in Group E, which is composed of sequences from tetraploid S. plicata (one sequence) and hexaploid S. palmifolia (3 sequences), indicating another novel genome in the Setaria species group. Four unique sequences were obtained from hexaploid S. arenaria and they were classified into Group F, which implies yet another new genome in the Setaria genus.

Phylogeny based on kn1sequences

Twenty-nine kn1 unique sequences were included in the phylogenetic analysis, obtained from the samples we used as well as one from S. palmifolia in GenBank [EF189834]. The consensus Bayesian tree in Figure 3 was similar in the groups that were delimited in the 5S tree (and similar results were found for the neighbor joining tree, although with less resolution). Group A was composed of the A genome sequences from 5 species, which include S. viridis, S. italica, S. verticillata, S. faberi and S. queenslandica. Group B was made up of 4 sequences corresponding with the known B genome from S. adhaerans, S. verticillata, and Qing 9, however, we did not detect any sequences from S. faberi in this group, albeit its tetraploid AABB genome constitution [19]. For the known C genome only two sequences were amplified from S. grisebachii and S. lachnea respectively, which form Group C in the tree constructed. However, one sequence from S. lachnea (W74-31) was aligned closely to the E group clade with high support. All sequences of kn1 from S. glauca and S. parviflora were classified into Group D, which agrees with the GISH result that S. glauca and S. parviflora share the same or similar genomes. Two sequences of kn1 from S. plicata, two sequences from S. palmifolia and the accession of S. palmifolia from GenBank [EF189834] were aligned into Group E, which also agrees with the GISH result and 5S rDNA tree. Only two unique sequences of kn1 were amplified from the hexaploid S. arenaria, which forms Group F in the phylogenetic tree, and their clear differences from other kn1 sequences is further evidence that the corresponding genomes in S. arenaria are distinct from the known A, B, C genomes and from the S. glauca and S. palmifolia genomes.

Differences in the relationships between the genome types are evident by comparison of the 5S rDNA and kn1 trees. In particular, the relationships of genome E (S. plicata, S. palmifolia) is strongly supported as being with genome B in the kn1 tree but with genome A and C (albeit with less support) in the 5S rDNA tree. Likewise, in the kn1 tree genomes A and D are strongly supported as a phylogenetic group whereas in the 5S rDNA tree genome D is sister to a clade of A, C, and E genome species. In both trees, genome F (S. arenaria) is basal.

Novel genomes identified by GISH and phylogenetic trees in the genus Setaria

There are over 125 species in the genus of Setaria[2], yet differences in genome constitution have been poorly studied. The genome of diploid species of green foxtail and foxtail millet was first designated as A genome by Li et al [10], and the genomes of diploid S. adhaerans and S. grisebachii were designated as B and C genomes respectively [19, 20]. The genome constitutions of two tetraploid species, S. verticillata and S. faberi, were identified being AABB by GISH [19], and that of S. queenslandica as AAAA using GISH in our lab [20]; S. queenslandica being the only autotetraploid found in the Setaria genus. So far no other reports on genomes or genome constitution of Setaria species have been published.

Previous reports by GISH identified that the genomes in S. glauca were neither A nor B [19], and our result in this report from both GISH and phylogenetic trees indicates that genomes in S. glauca were also quite distinct from the newly identified C genome, suggesting a new genome in S. glauca which we designated as genome D. A close relationship between S. glauca and S. parviflora was identified by GISH and phylogenetic trees, imply that they share the D genome. Little or no signals were detected when chromosomes of S. plicata and S. palmifolia were hybridized with probes from genomic DNA of genomes A, B, C and that of S. glauca, and this is supported by the phylogenetic analyses that also show that the sequences from S. plicata and S. palmifolia are clearly distinguished from those of the known genomes and that of S. glauca and S. parviflora. This indicates another new genome in Setaria, designated as genome E, shared by S. plicata and S. palmifolia. The GISH pattern and phylogenetic result also indicate that the three sets of genomes in S. arenaria are distantly related with the known genome of A, B, C and that of S. glauca and S. plicata, suggesting another new genome in S. arenaria, designated genome F. Thus three new genomes are suggested by combined data from GISH and phylogenetic analyses, although the corresponding species of these new genomes are all polyploid. So far six distinct genome-types coexist in the genus of Setaria, reflecting the diversified and diverged genetic composition of this group of grasses, and in good concordance with ISSR result [13] and phylogenetic analyses [8, 9]. Setaria is suspected to be polyphyletic, based on phylogenetic analysis using both chloroplast and nuclear markers [8, 9], so it will be interesting to know whether related taxa outside the various groups of Setaria studied here share these genome types.

The B genome in S. faberi is distinct from the B genome in S. verticillata

S. verticillata and S. faberi were all identified as being allotetraploids with a genome constitution of AABB [19]. However, the two primer pairs amplifying the 5S rDNA and kn1 gene fragments only amplified from both the B genome diploid of S. adhaerans and the B genome of allotetraploid of S. verticillata, but gave no amplification from the putative B genome in S. faberi (Figures 2 & 3), even though multiple PCR amplifications were tried and multiple clones sequenced. The same results were also obtained by Benabdelmouna et al [14] for the 5S rDNA marker. This suggests that the B genome in S. faberi has either diverged from the B genome of S. adhaerans and S. verticillata, or further recombination or backcrossing has occurred that led to the loss of those gene copies. Analysis of more molecular markers and a greater number of accessions will be necessary to distinguish between these possibilities. S. adhaerans and S. verticillata are both characterized by bristles with retrorse barbs, but the bristles of S. faberi are the common straight ones. S. verticillata appears to have originated from genome duplication of a hybrid between an A genome diploid and a B genome diploid, and the B genome diploid is genetically closely related with the current S. adhaerans. The origin of S. faberi was from genome duplication of a hybrid between an A genome diploid and an unknown diploid that are relatively distantly related to S. adhaerans B genome. Previous hypotheses have suggested that S. faberi originated from S. italica x S. adhaerens, based on the presence of large seeds [43]. Our results do not specifically address this hypothesis, but we suspect that large seeds in S. italica are the result of human selection during the last 10,000 years, and that S. faberi is of older origin than this. However, genetic diversity analysis of S. faberi accessions show little genetic variation [44], indicating that this taxon requires more study.

Genome constitution of S. lachneawas probably CCC’C’

Our GISH experiments clearly show that the two sets of chromosomes of S. lachnea had strong hybridization signals covering the entire chromosomes with probe from genomic DNA of diploid S. grisebachii, suggesting that S. lachnea is probably an autotetroploid species with genome constitution of CCCC (Figure 1i-j). To confirm this conjecture, multiple independent hybridizations with different washing protocol with higher stringency were carried out, giving the same results, and no differences between the two sets of chromosomes were detected (Additional file 2: Figure S1). However, only one of the two unique sequences of kn1 from S. lachnea was closely related to S. grisebachii representing the C genome, while the other (W74-31) was more closely related to the E genome sequences (Figure 3). We did not test hybridization of E genome probes to S. lachnea, so it remains a possibility that the second genome is related to the E genome. The evidence suggests that one genome set of S. lachnea is C, and the other set is a genome closely related with C with some sequence divergence, thus we describe the genome constitution of S. lachnea as CCC’C’.

Qing 9 is probably a newly identified species in B genome of Setaria

Qing 9 is a Setaria accession collected from Qiema of Luancheng, Shijiazhuang, Hebei province in China in 2003, and treated as a S. viridis sample. However, its morphology is clearly distinct from that of S. viridis, and ISSR data also indicated that it was distantly related with S. viridis samples with A genome, but closely related with B genome species S. adhaerans[13]. Our GISH results show that the chromosomes of Qing 9 hybridized well with probe from the B genome S. adhaerans, whereas no signal was detected when using the A genome S. viridis as probe. This clearly indicated that it is a B genome species and that it was misclassified as a S. viridis sample. The 5S rDNA and kn1 gene phylogenetic trees also clearly show that Qing 9 is closely related with S. adhaerans and distantly related with S. viridis, which well support our conjecture. So, Qing 9 is clearly a B genome diploid sample.

Morphological observations also clearly distinguishes Qing 9 from S. adhaerans. As a typical B genome species, S. adhaerans is characterized with bristles of deep retrorse barbs, tall and slim stem, and narrow leaf blade. But Qing 9 is characterized with shorter and sturdy stem, wide leaf blade, and especially its bristles are weakly retrorse hooked, which is not only clearly different from the deep retrorse hooked S. adhaerans but also clearly different from the straight ones of S. viridis. Qing 9 does not correspond to any of the twenty three indigenous species described in the Flora of China [45], suggesting that it is a newly identified species. Combine all those data, we make the hypothesis that Qing 9 is probably a not yet detected resident species in China. S. adhaerans is a sub-tropical B genome diploid species found around the Mediterranean area (including South Europe and North Africa), thus the identification of novel B genome diploid species indigenous to China will be important for the evolutionary study of the Setaria group and the grass family. Detailed studies of this taxon are currently under way.