Triparental origin of triploid onion, Allium × cornutum (Clementi ex Visiani, 1842), as evidenced by molecular, phylogenetic and cytogenetic analyses

Molecular phylogenetic analysis of ITS and NTS sequences

The sequences of the ITS regions of A. × cornutum were aligned to other ITS sequence data for Allium species of section Cepa available in GenBank (see Table 2 for the details). Three distinct clades of ITS1-5.8S-ITS2 sequences were identified within the A. × cornutum genome (Figure 1a, Additional file 1: Figure S1). Most of the sequences (41 sequences) formed one large clade, which also included ITS sequences of the wild Asian species A. pskemense and has therefore been designated xas clade P (“pskemense”-type). A second clade, containing six ITS sequences of A. × cornutum, grouped with the sequences of A. roylei and has been designated as clade R (“roylei”-type). One sequence of A. ×cornutum (from an individual from Hvar) showed similarity to A. cepa and A. vavilovii sequences, and hence, the whole clade was designated as clade C (“cepa”-type) (Figure 1a). The seven A. × cornutum ITS sequences consisting of clades R and C had a clearly distinguishable 13 base (CTGTAAACATACT) insertion in the ITS2 region, which is shared by both A. cepa and A. roylei but absent in A. pskemense (Additional file 1: Figure S1).

	GenBank accession number (reference)
	ITS1-5.8S-ITS2		NTS-5S
Allium pskemense	AM418380	(Gurushidze et al. 2007) [22]	JF496621	(Son et al. 2012) [28]
	AM418382	(Gurushidze et al. 2007)	JF496622	(Son et al. 2012)
	AJ411907	(Friesen et al. 2006) [23]
Alllium roylei	AJ411945	(Friesen et al. 2006)	KC731587	This study
	AM492189	(Gurushidze et al. 2007)	KC731590	This study
Allium cepa	FJ664287	(Hirschegger et al. 2010) [24]	AB056584	(Shibata and Hizume 2002) [29]
	AM418367	(Gurushidze et al. 2007)	AB056593	(Shibata and Hizume 2002)
	AM418370	(Gurushidze et al. 2007)
Allium vavilovii	AM418383	(Gurushidze et al. 2007)	JF496618	(Son et al. 2012)
			JF496619	(Son et al. 2012)
			JF496620	(Son et al. 2012)
Allium cepa var. aggregatum			JF496648	(Son et al. 2012)
Allium oschaninii	AM418376	(Gurushidze et al. 2007)
Allium praemixtum	AM418379	(Gurushidze et al. 2007)
Allium farctum	AM492184	(Gurushidze et al. 2007)
Allium asarense	AM418365	(Gurushidze et al. 2007)
Allium altaicum	GQ412198 GQ181094	(Jang et al. unpublished) (Li et al. 2010) [25]	JF496602	(Son et al. 2012)
Allium fistulosum	JF990845	(Guenaoui et al. 2013) [26]	JF496610	(Son et al. 2012)
Allium x cepiforme	GU566611	(Li et al. 2010)
Allium galanthum	GQ181101	(Li et al. 2010)
Allium x proliferum			JF496645	(Son et al. 2012)
Allium schoenoprasum	AY427547	(Ricroch et .al. 2005) [27]	AB066483	(Shibata and Hizume 2002)
	GQ412234	Jang et al. unpublished	AB066482	(Shibata and Hizume 2002)
			AB066474	(Shibata and Hizume 2002)
Allium maximowiczii	GQ412215	Jang et al. unpublished
Allium deltoidefistulosum	GQ412203	Jang et al. unpublished
Allium linearifolium	GQ412206	Jang et al. unpublished
Allium thunbergii	GQ412255	Jang et al. unpublished
Allium condensatum	GQ412201	Jang et al. unpublished

The exceptions are the sequences for the NTS-5S genes of A. roylei, which were obtained in the present study.

The presence of different ITS1-5.8S-ITS2 rDNA repeat types within the A. × cornutum genome and their genetic similarity to their putative parental species A. pskemense, A. roylei, and A. cepa was confirmed by phylogenetic analysis. The phylogenetic algorithms Maximum-Likelihood (ML) and Bayesian Inference (BI) resulted in nearly the same tree topology. Only the BI tree is shown (Figure 1a), which summarises the topology and posterior probabilities (PP) from BI plus bootstrap support (BS) from the ML analysis. All the ITS sequences of A. × cornutum clustered into the three clades: P, R and C (Figure 1a). The clades P and R were both strongly supported by both analyses (clade P: 1.0 PP, 98% BS; clade R: 0.98 PP, 97% BS). The clade C containing a single A. × cornutum sequence (Hvar 21) and sequences from A. cepa, A. vavilovii and A. asarense was less strongly supported (1.0 PP; 69% BS).

To analyse the variability of non-transcribed spacer (NTS) of the 5S rRNA genes in A. × cornutum, 19 clones were obtained from three A. × cornutum individuals (Table 1). The NTS region showed higher sequence variation than the ITS1 and 2 regions. Nearly all of the cloned sequences were unique. The length of the 5S rDNA unit including the non-transcribed spacer (NTS) region ranged from 339 to 346 bps. The conserved 5S rRNA coding region was excluded from further analyses. The NTS region comprised 224–231 characters, of which 148 were constant and 73 were variable (including 53 parsimony uninformative and 20 parsimony-informative characters). In total, 13 positions in the alignment included gaps (Additional file 2: Figure S2).

Phylogenetic analyses of NTS clones of A. × cornutum together with NTS sequences of their potential close relatives retrieved from GenBank (Table 2) supported the grouping of the A. × cornutum NTS sequences into the three well-supported clades (P, R and C; Figure 1b). The largest clade consisted of 11 sequences and grouped with the NTS sequences of Allium pskemense, A. altaicum and A. × proliferum (88–92% similarity), and this clade has been designated as clade P. The second largest clade (R) consisted of five NTS sequences of A. × cornutum, which grouped with A. roylei NTS regions (89% similarity). The smallest clade (C) consisted of only three NTS sequences of A. × cornutum, which were 99% identical to the NTS sequences of A. cepa and A. vavilovii. The largest clade P, comprising 11 NTS sequences of A. × cornutum, has been clearly separated from the NTS sequences of its closest relatives, A. altaicum, A. ×proliferum, and A. pskemense (Figure 1b).

Chromosomal localisation of 5S rDNA and 35S rDNA genes and genomic in situhybridisation

The number and localisation of 5S and 35S rDNA loci were analysed in two clones of Allium × cornutum (Dubrava and Vis) and in three putative parental diploid taxa that were identified in phylogenetic analyses: Allium cepa, A. pskemense and A. roylei. Allium cepa possessed two loci of 35S rDNA in two satellite chromosome pairs and two 5S rDNA loci, both localised on the short arm of chromosome 7 (Figure 2a, b). The 35S rDNA loci differed in the intensity (and likely number of repeat copies). A. pskemense possessed one subterminal locus of 35S rDNA located on one satellite chromosome pair and at most two loci of 5S rDNA on chromosome 6 (Figure 2c, d). One larger locus of 5S rDNA was located interstitially in the short arm of chromosome 6 and the other, weaker locus was situated in the pericentric region of the long arm of the same chromosome. Two homologous chromosomes carrying 5S rDNA exhibited heterozygosity concerning the presence of a small 5S rDNA signal in the pericentromeric region of the long chromosome arm (Figure 2c,d). A. roylei possessed a single locus of 35S rDNA on a satellite-bearing chromosome pair (Figure 2e, f) and two 5S rDNA loci of similar size in the small metacentric chromosome 7. One of these 5S rDNA loci was localised close to the pericentric region, and the other was localised more internally within the same arm (Figure 2e,f).

In triploid A. × cornutum, two major subterminally localised signals of 35S rDNA were detected in the short arms of the two subtelocentric satellite chromosomes. A third minor 35S rDNA signal was detected on some spreads and was located in the subtelomeric region of one small sub-metacentric chromosome (white arrows in Figure 2g, h, i). The largest satellite chromosome (resembling the large NOR-bearing chromosome of A. cepa) lacked a 35S rDNA signal. Each of the three 35S rDNA signals differed in intensity and size, with the medium-sized chromosome carrying the strongest signal.

The 5S rRNA genes were detected in three differently sized chromosomes in A. × cornutum (Figure 2g, h, i). The largest chromosome carrying a 5S rDNA signal resembled the typical A. cepa chromosome with two 5S rDNA loci localised within the long arm of chromosome 7. The medium sized chromosome carrying 5S rDNA in A. × cornutum possessed two 5S rDNA loci, which differed in size and intensity, with the stronger signal positioned interstitially within the short arm and the weaker signal corresponding to the pericentromeric region of the long chromosome arm. This chromosome bears resemblance to the 5S rDNA-bearing A. pskemense chromosome. The smallest chromosome carrying 5S rDNA in triploid Allium possessed only one signal in the pericentromeric region of the short chromosome arm (Figure 2g, h, i). This chromosome might represent a truncated and/or rearranged chromosome 7 of A. roylei (Figure 2e, f).

The results of FISH mapping of 5S rDNA supported the inferences of the phylogenetic analyses of the 5S and 35S rDNA of triparental origin of A. × cornutum triploids. The three chromosomes carrying 5S rDNA genes in A. × cornutum likely originated from three different diploid Allium species (Figure 3).

A combination of genomic in situ hybridisation (GISH) and FISH was attempted to identify the genomic origin of 5S rDNA-bearing chromosomes (Figures 4a–d; 5b, c). A single chromosome carrying two 5S rDNA loci (designated as P) was clearly labelled with A. pskemense genomic DNA, in contrast to the two other chromosomes carrying the 5S rDNA signals (designated as R and C), which remained unlabelled or were only weakly labelled (Figure 4b). The hybridisation of the genomic DNA of A. roylei to the incomplete metaphase plate of A. × cornutum in the presence of an excess of unlabelled A. cepa total genomic DNA as blocking DNA (Figure 4c) allowed the labelling of six chromosomes of the triploid onion, which are marked by arrows. Among the chromosomes labelled with A. roylei genomic DNA, one chromosome carrying 5S rDNA is indicated (designated as R; Figure 4d). This chromosome corresponds to the smallest 5S rDNA-bearing chromosome, which is likely a truncated A. roylei-originating chromosome (Figure 3). With the aim to simultaneously discriminate between the three genomes of A. × cornutum (the putative C, R and P genomes), we performed genomic in situ hybridisation with two labelled parental DNA sequences as probes (A. pskemense genomic DNA was labelled in green; A. roylei genomic DNA was labelled in red), whereas the genomic DNA of A. cepa remained unlabelled and was used as blocking DNA (Figure 5a, b). Eight chromosomes were labelled predominantly with A. pskemense genomic DNA (P-genome; green; Figure 5a); furthermore, eight chromosomes hybridised predominantly with A. roylei genomic DNA (R-genome; red), and eight chromosomes remained nearly unlabelled (C-genome, blue). The chromosomes have subsequently been reprobed with 35S and 5S rDNA probes. A comparison of Figure 5b and c confirms that the three different 5S-bearing chromosomes of A. × cornutum belong to three different parental genomes.

Triparental allopolyploid origin of A. × cornutumand the identification of its putative parental species

Two contradictory studies have been published concerning the origin of the triploid onion A. × cornutum [6,11; reviewed in 12]. The suggested triparental origin of a triploid has been proposed, but only two putative parental taxa were suggested (A. cepa and A. roylei), with the third parental taxon remaining unknown (the so-called “X genome”; [6]). The present study provides phylogenetic evidence for a triparental hybrid origin of A. × cornutum, supported by the mapping of the 5S and 35S rRNA genes in the chromosomes of putative parental taxa and the hybrid. Additionally, the third putative parental species, the wild Asian species A. pskemense B. Fedtsh., has been identified. These data support the complex hybrid origin of A. × cornutum and allow the rejection of previous hypotheses [11], which postulated the origin of A. × cornutum as a derivative of A. cepa. The current study clearly demonstrates that A. × cornutum contains three types of both ITS and NTS sequences, each grouping with one of three putative parental taxa, Allium pskemense, A. roylei, and A. cepa.

GISH allowed the detection of the three parental genomes in the hybrid, despite some level of cross-hybridisation. Using a combination of GISH and FISH, 5S and 35S rDNA-bearing chromosomes of the hybrid were shown to originate from the respective chromosomes of the putative parents.

ITS sequence variability and 35S rDNA mapping

The phylogenetic analysis of the ITS region of A. × cornutum revealed three major ITS types denoted as P, R, and C. These types were recovered as separate clades in combined analyses and were shown to bear high sequence similarity to three diploid Allium species/lineages: A. pskemense, A. roylei, and A. cepa/A. vavilovii. A. cepa and A. vavilovii are closely related, with A. cepa being known only as a cultivated taxon; A. vavilovii was inferred as its closest wild relative [22].

Concordantly, three 35S rDNA loci were detected on three different chromosomes in triploid onion, in agreement with a previous report [52], and could be assigned to the three putative parental genomes. Our previous GISH analysis [6] showed that the medium-sized NOR chromosome carrying the largest 35S rDNA locus did not hybridise either with genomic probes of A. cepa or of A. roylei, thus being assigned to the unidentified X genome. The current study showed that this locus originates from A. pskemense and was the source of the majority of the cloned ITS sequence regions (41 out of 48 ITS clones). The major 35S rDNA locus originating from the A. cepa genome has either been lost during the evolution of the triploid genome or contains only very few copies that are below the detection limit of FISH. The second (minor) A. cepa locus, which has been detected in the triploid hybrid on the medium-sized submetacentric chromosome, contains a very small copy number and might be in the process of being lost. The single ITS sequence of the C-type (clone Hvar 21) that was recovered from the triploid onion genome likely represents the A. cepa minor 35S rDNA locus.

Earlier analyses of the activity of the NOR regions in triploid onion using silver staining indicated that all three 35S rDNA loci were active with a maximum of three nucleoli detected in the interphase nuclei of all five Croatian clones of A. × cornutum[14]. In contrast, Pran, the Indian clone of triploid A. × cornutum, possessed only a single active NOR on a medium–sized satellite chromosome [14]. Such a result indicates ongoing evolution of the rDNA in triploid onions over the whole species distribution range. Multiple origins of this triploid hybrid taxon are currently excluded based on the unique genome size, isozyme, RAPD and RFLP patterns of Pran, Ljutika and other analysed clones of triploid onion [6, 10, 11].

NTS sequence variation and 5S rDNA mapping

The divergence of the NTS sequences in several well-documented allopolyploid systems proved very useful for the identification of the putative parent species: i.e., Nicotiana tabacum[48], Zingeria[53], Anemone multifida and A. baldensis[47], and Melampodium[1]. The 5S rDNA NTS sequences of A. × cornutum clustered into the three main clades (C, P, and R), which had high sequence homology to the three putative parental species A. cepa, A. pskemense, and A. roylei, respectively. Whereas the clades C (A. cepa) and R (A. roylei) were well supported, the clade P (A. pskemense) failed to form a single well-supported clade with its closest relatives, A. pskemense, A. altaicum, and A. × proliferum. This result could have been caused by significant intra-individual variability within the NTS region [28] as well as possible high genetic variation of 5S rDNA within A. pskemense and its relatives across their wide geographical distribution.

The results of the NTS sequence analysis were largely congruent with the cytogenetic mapping of 5S rDNA in A. × cornutum, and 5S rDNA loci were detected in different positions in three chromosomes of different morphology. The two larger 5S rDNA-bearing chromosomes greatly resembled the putative parental chromosomes of A. cepa and A. pskemense. The 5S rDNA-bearing chromosome having 5S rDNA loci distributed in a similar manner as in A. pskemense has also been labelled with A. pskemense genomic DNA in a GISH experiment. The smallest 5S rDNA-bearing chromosome of A. × cornutum has been shown to hybridise with genomic DNA of A. roylei ([6]; current study), thus confirming its origin from the R genome (A. roylei). This chromosome, however, differed from the parental 5S rDNA-bearing chromosome of A. roylei and possessed only a single 5S rDNA locus in the pericentromeric region of the short arm of the chromosome instead of two loci. At least two different scenarios can account for the observed truncation: i) the smallest 5S rDNA-bearing chromosome and the entire R genome could have originated from a diploid ancestor closely related to A. roylei, which is characterised by smaller chromosomes and only a single 5S rDNA locus; and ii) chromosomes of A. × cornutum originating from the R genome have undergone rearrangements after hybridisation and lost one of the 5S rDNA signals in the process. Based on our current understanding of genome restructuring after allopolyploidisation and a general trend of diploidisation of both the 5S and 35S rDNA loci, the second hypothesis is more likely [43, 54–57], especially because the sequences of A. × cornutum in clade “R” are very similar to those of A. roylei. The triploid genome of A. × cornutum might have been subjected to additional genomic rearrangements such as inter-chromosomal translocation, deletions, and/or inversions.

FISH mapping of 35S and 5S rRNA genes in A. pskemense and A. roylei and inferences of the phylogenetic relationships in section Cepa

Although the phylogenetic relationships among 12 species of sections of Cepa have been inferred from analyses of plastid regions and ITS [20, 22], the chromosomal positions of 35S and 5S rDNA have so far been determined for only three species: A. cepa, A. fistulosum and A. altaicum[58–61]. The two 5S rDNA loci in A. cepa were located on the longer arm of chromosome 7 (as confirmed by our data), whereas in A. fistulosum a single 5S rDNA locus was detected interstitially in the short arm of chromosome 7. In natural diploid homoploid hybrids of A. cepa and A. fistulosm, top onions (Allium × proliferum (Moench) Schrad. and Allium wakegi Araki (both 2n = 16), the two chromosomes carrying the 5S rDNA signals corresponded to the 5S rDNA-bearing chromosomes of the parental species [58]. In this study, 35S and 5S ribosomal genes were mapped for the first time in somatic chromosomes of diploid A. pskemense and A. roylei. These two species possessed an identical number of 35S rDNA loci but differed in the positions of the 5S rDNA loci on chromosome 7. The localisation of 5S rDNA in A. roylei chromosome 7 resembles more closely that of A. cepa, whereas A. pskemense is more similar to A. fistulosum and A. altaicum. This finding supports the hypothesis of Son et al.[28] that A. pskemense is more closely related to A. fistulosum and A. altaicum. The number and position of the 5S rDNA loci, therefore, proved to be evolutionarily informative in analysing the species of Allium from section Cepa.

Plant materials and DNA extraction

Four clones of A. × cornutum (known in Croatia under the name Ljutika) were obtained from local gardens and vineyards at four well-separated localities of the Croatian seaside region (Dubrava and Kaštela) and islands (Vis, Hvar). A. pskemense B. Fedt. (CGN21442) and A. roylei Stearn (CGN20520) seeds were kindly provided by the Centre for Plant Breeding and Reproduction Research, Wageningen, The Netherlands. The commercial cultivar A. cepa cv, ‘Holland Yellow’ was used to obtain the DNA and chromosome complements of A. cepa Genomic DNA was extracted from young leaves using the CTAB method according to Saghai Maaroof et al.[62].

PCR amplification and cloning

The ITS1-5.8S-ITS2 region of 35S rDNA was amplified by PCR using the universal primers ITS1 and ITS4 and the procedures described by Bezić et al.[63]. The whole coding and non-transcribed spacer (NTS) region of the 5S rDNA gene was amplified using the primers and conditions from Weiss-Schneeweiss et al.[57]. The amplified products were visualised and confirmed by 1% agarose gel electrophoresis, extracted from the gel, ligated into pGEM-T Easy vectors (Promega, Madison, Wisconsin, USA) and cloned into competent JM109 E. coli cells. DNA from individual plasmids carrying inserts was isolated using a Plasmid Mini Kit (Qiagen, Hilden, Germany). Purified plasmid DNA was sent to Macrogen (Seoul, Korea) for sequencing of the inserts.

Sequence analysis

The DNA sequences were assembled and prealigned using BioEdit ver. 7.0.5.3 [64]. They were then aligned in ClustalW [65] and implemented in MEGA5 [66], and the alignment was refined manually. The sequences were deposited in GenBank (Table 1). To avoid multiple submissions of identical sequences, we sent only one sequence of each type. To infer the phylogenetic relationships from the newly obtained ITS and NTS sequences of A. × cornutum and other closely related Allium species, the sequences were subjected to a similarity search against the non-redundant nucleotide sequence database using the NCBI (National Centre for Biotechnology Information) BLASTN network service. Sequence alignments of newly amplified regions and sequences of other related Allium species deposited in GenBank were performed using MEGA5 [66]. Polymorphic and variable sites as well as different haplotypes were generated using DnaSP Ver. 5.10 [67]. A Bayesian analysis was performed with MrBayes 3.1 [68] with 4 chains of 1,000,000 generations, trees sampled every 100 generations and the burn-in value set to 25% of the sampled trees. The best-fit substitution model was used as determined by the Akaike Information Criterion [69] as implemented in jModelTest 0.1.1 [70]. A maximum-likelihood analysis using starting trees obtained by neighbour-joining and TBR branch swapping with model parameters was performed using PAUP* 4.0b10 [71]. The number of bootstrap replicates was set to 1000. Phylogenetic trees were displayed in FigTree v1.3.1.

Chromosome preparation, fluorescence in situ hybridisation (FISH) and genomic in situhybridisation (GISH)

Chromosomes for FISH and GISH were prepared as described by Puizina et al.[6]. Clone pTa794 contained the complete 410-bp BamHI fragment of the 5S rRNA gene, and the spacer region of wheat [72] was used as the 5S rDNA probe. The 2.4 kb HindIII fragment of the partial 18S rRNA gene and ITS1 from Cucurbita pepo, cloned into pUC19 [73], were used as the 18S rDNA probe. The 5S rDNA probe was labelled with digoxygenin using a DIG-nick translation kit (Roche Diagnostics, Mannheim, Germany), whereas 18S rDNA was labelled with biotin using a BIO-nick translation kit (Roche Diagnostics, Mannheim, Germany). The genomic DNA (1 μg/reaction) was labelled with biotin using a BIO-nick translation kit (Roche Diagnostics, Mannheim, Germany) according to the supplier’s instructions.

The FISH method followed the procedures outlined in Weiss-Schneeweiss et al.[57, 74]. Briefly, the preparations were re-fixed and air dried, and chromosomal DNA was denatured in 70% (v/v) deionised formamide in 2× SSC, pH 7.0 at 70°C for 2 min, dehydrated through an ethanol series and air-dried. The hybridisation mixture containing labelled probes (100–150 ng/slide), 20–100x excess blocking DNA (for GISH) or salmon sperm DNA (for FISH), 50% formamide, 2x SSC, 10% dextran sulphate, and 0.15% sodium dodecyl sulphate (SDS) was denatured at 75°C for 10 min. The probe was applied to the slides and incubated at 37°C in a humid chamber overnight. Subsequently, the slides were washed for 5 min in 2x SSC at 39°C, for 5 min in 0.1x SSC at 39°C, and for 5 min in 2x SSC + 0.2% Tween 20 at 39°C. Biotin- and digoxygenin-labelled probes were detected using Extravidin-Cy3 (2.5 μg/mL) and anti-digoxygenin-FITC (5 μg/mL), respectively, both in 2% BSA in 2x SSC + 0.2% Tween 20 buffer at 37°C for 60 min. The slides were subsequently washed twice for 7 min in 2x SSC at 42°C and for 7 min in 2x SSC + 0.2% Tween 20 at 42°C. Finally, they were mounted in 20 μL of the antifade solution Vectashield containing 0.5 μg/mL DAPI (Vector Laboratories, Burlingame, CA, USA) and stored at 4°C. The slides were examined with a Zeiss Axioimager M1 epifluorescence microscope with a high-resolution microscopy camera (Carl Zeiss AxioCam MR Rev3) using Axio Vision Rel. 4.7 software (Karl Zeiss, Vienna, Austria). For rDNA localisation, an average of 15–20 metaphase plates were analysed for each species. GISH hybridisation and detection were performed using the same protocols.