The origin of Cardamine flexuosa (Wavy Bittercress) has been a conundrum for more than six decades. Here we identify its parental species, analyse its genome structure in comparison to parental genomes and describe intergenomic structural variations in C. flexuosa.
Genomic in situ hybridization (GISH) and comparative chromosome painting (CCP) uncovered the parental genomes and the chromosome composition of C. flexuosa and its presumed diploid progenitors.
Cardamine flexuosa is an allotetraploid (2n = 4x = 32), originating from two diploid species, Cardamine amara and Cardamine hirsuta (2n = 2x = 16). The two parental species display almost perfectly conserved chromosomal collinearity for seven out of the eight chromosomes. A 13 Mb pericentric inversion distinguishes chromosome CA1 from CH1.
A comparative cytomolecular map was established for C. flexuosa by CCP/GISH. Whereas conserved chromosome collinearity between the C. amara and C. hirsuta subgenomes might have promoted intergenomic rearrangements through homeologous recombination, only one reciprocal translocation between two homeologues has occurred since the origin of C. flexuosa. The genome of C. flexuosa demonstrates that allopolyploids can maintain remarkably stable subgenomes over 104–105 yr throughout a wide distribution range. By contrast, the rRNA genes underwent genome-specific elimination towards a diploid-like number of loci.
Hybridization and genome doubling may explain the saltatory origin of plant species. All seed plants have undergone at least one round of whole-genome duplication (WGD; Jiao et al., 2011) and multiple rounds of WGD are particularly characteristic of angiosperms (Leitch & Leitch, 2012). Both recurrent polyploidization events and gradual genome diploidization (Mandáková et al., 2010) influenced the make-up of extant plant genomes. Whereas recent WGDs are self-evident, and parental genomes of autopolyploid or allopolyploid species can usually be pinpointed, ancient polyploidization events are more difficult to reconstruct in quasidiploid genomes after millions of yr of evolution. Regardless of the time of occurrence, each WGD increases the complexity and redundancy of plant genomes.
The complex structure of polyploid genomes results from a merger of two or more genomes, manifested by increased genome size and chromosome number. Multiple sets of homologous chromosomes in autopolyploids, and homologous as well as homeologous chromosomes in allopolyploids, make meiotic chromosome pairing and segregation more demanding to regulate. The increased genomic complexity of allopolyploid genomes is also reflected by increased amounts of dispersed and tandem repeats, as well as by multiple copies of homeologous genes, whose expression is regulated epigenetically (Birchler & Veitia, 2012; Madlung & Wendel, 2013). Sequence similarity of homeologous genes and repeated sequences hampers the dissection of an allopolyploid genome into its subgenome components. Many repeats are common between the two or more hybridizing genomes as a result of shared ancestry. In addition, genome-specific repeats are homogenized by intergenomic ectopic recombination (Lim et al., 2007a; Luo et al., 2010) and rates of meiotic recombination are significantly higher in polyploids than in diploids (Pecinka et al., 2011). As a consequence, genomic analyses and mapping in polyploid genomes are impeded by the problematic assignment of individual sequences, markers or gene copies to parental (sub)genomes.
The structure of allopolyploid genomes can be reconstructed using a bottom-up approach by analysis of the original diploid genomes and by identification of parental genomic features within the polyploid genome. Alternatively, the polyploid genome can be analyzed as a whole, and its reconstructed subgenome components can be compared with the genomes of the purported parents. Historically, several methodological approaches have been used for mapping allopolyploid genomes up to contemporary integrated genome analyses. Genetic maps constructed using different types of markers are available for allopolyploid crops such as durum wheat (Triticum durum Desf.), rapeseed (Brassica napus L.) or tetraploid cottons (Gossypium barbadense L. and G. hirsutum L.; Mantovani et al., 2008; Wang et al., 2011a; Blenda et al., 2012). The advent of massive parallel sequencing allowed for whole-genome shotgun sequencing of an increasing number of polyploid crop species (Wang et al., 2011b; Potato Genome Sequencing Consortium, 2011; Paterson et al., 2012). However, owing to the complexity of polyploid genomes, it remains challenging to assign short read sequences to individual chromosomes and to two or more subgenomes. To overcome these difficulties, bacterial artificial chromosome (BAC)-based physical maps combined with high-density genetic maps were used to improve the accuracy of whole-genome sequence assemblies (Luo et al., 2010; Ariyadasa & Stein, 2012; Sierro et al., 2013). Comparative chromosome painting is a method of choice for genome mapping in allopolyploid plant species in families with small genome sizes and available genomics resources (e.g. chromosome-specific BAC or fosmid libraries, comparative genetic maps, etc.).
Wavy Bittercress (Cardamine flexuosa With.) is a tetraploid species (2n = 4x = 32) distributed throughout the European continent, from whence it was introduced to North America. Published data on the occurrence of C. flexuosa from Asia, Australia, Africa and mostly also from the Americas refer to populations that should be treated as a different species (octoploid of 2n = 8x = 64, having the most likely different parentage compared with European tetraploids; Lihová et al., 2006) that is currently informally treated as ‘Asian C. flexuosa’, pending further studies (Lihová et al., 2006; Al-Shehbaz et al., 2010). European C. flexuosa was considered as an autopolyploid (Banach, 1950) or allopolyploid species (Ellis & Jones, 1969). Whereas Cardamine hirsuta L. (2n = 2x = 16) was considered very likely to be one of the parental species, based on the close morphological resemblance to C. flexuosa, the identity of a second genome donor remained elusive for a long time. Morphological similarities between the diploid C. hirsuta and C. flexuosa are so profound that both species were often misidentified and confused with each other. Nevertheless, there are several discriminating morphological characters that permit correct identification of both species (Marhold, 1995). Close morphological affinities between C. hirsuta and C. flexuosa led Banach (1950) to describe the latter species as an autotetraploid derivative of the diploid C. hirsuta. Based on morphometric measurements, Ellis & Jones (1969) put forward C. hirsuta and Cardamine impatiens (2n = 2x = 16) as the putative parents of C. flexuosa, and the diploid European Cardamine amara L. ssp. amara (2n = 2x = 16) was hypothesized as a probable maternal parent by Lihová et al. (2006). Thus, although European C. flexuosa is a widespread species, its origin and genome evolution have remained unclear until recently.
Here we present, for the first time, a whole-genome cytomolecular map of an allopolyploid plant. By combining comparative chromosome painting and genomic in situ hybridization (CCP/GISH), we elucidate the origin and evolution of the C. flexuosa chromosome complement. GISH analysis first revealed two diploid species, C. amara and C. hirsuta, as parental species of the allotetraploid C. flexuosa. Comparative whole-genome maps of C. amara and C. hirsuta were constructed using the proposed Ancestral Crucifer Karyotype (ACK; Schranz et al., 2006), comprising eight chromosomes and 24 conserved genomic blocks (A to X), as the most likely ancestral genome of the genus Cardamine L. (Mandáková et al., 2013). Subsequently, the structure and evolutionary stability of both parental (sub)genomes within the allopolyploid genome of C. flexuosa were explored using CCP/GISH and in situ localization of rRNA genes. Furthermore, we tested the degree of conservation of the subgenome complements of C. flexuosa between populations across the European distribution area.
Materials and Methods
The following populations of C. amara, C. hirsuta, C. impatiens, Cardamine parviflora and C. flexuosa were analysed in this study: C. amara L. ssp. amara: Urnerboden (Switzerland; 46°53′14.9″N, 8°54′2.3″E), C. amara ssp. opicii (J. Presl & C. Presl) Čelak.: Mt Salatín (Slovakia; 48°57′29.8″N, 19°23′38.5″E), C. amara ssp. balcanica Marhold et al.: Mt Vitosha (Bulgaria; 42°34′54.7″N, 23°17′32.5″E); C. hirsuta L.: the reference Oxford accession (UK; Hay & Tsiantis, 2006), Gerhausen (Germany; 48°24.131′N, 9°49.419′E), Puerto de Lizarraga (Spain; 42°51.612′N, 2°00.605′E); C. impatiens L.: Kremnica (Slovakia; 48°42′26.1″N, 18°56′22.62″E); C. parviflora L.: Břeclav (Czech Republic; 48°56′20.79″N, 17°14′15.57″E); and C. flexuosa With. (Supporting Information Fig. S1): #1 Jormlien (Sweden; 64°44′10.9″N, 13°55′47.7″E), #2 Oppsal (Norway; 63°11′37.1″N, 8°52′37.6″E), #3 Sligo (Ireland; 54°20′34.8″N, 8°22′44.1″W), #4 Osnabrück (Germany; 52°17′7.43″N, 7°59′43.3″E), #5 Cikháj (Czech Republic; 49°38′50.103″N,15°57′27.724″E), #6 Zverovka (Slovakia; 49°14′54.6″N, 19°42′41.2″E), #7 Železné (Slovakia; 48°57′42.7″N, 19°24′17.5″E), and #8 Urnerboden (Switzerland; 46°53′14.9″N, 8°54′2.3″E).
Seedlings were either collected in the field or grown from seed in a growth chamber until bolting. Whole young inflorescences from different individuals within a sampled population were fixed in freshly prepared ethanol : acetic acid (3 : 1) overnight, transferred into 70% ethanol and stored at −20°C until use.
Chromosome spreads for GISH and CCP were prepared from anthers of young flower buds. Selected flower buds were first rinsed in distilled water and citrate buffer (10 mM sodium citrate, pH 4.8), digested by a 0.3% mix of pectolytic enzymes (cellulase, cytohelisase, pectolyase; all from Sigma) in citrate buffer for c. 3 h and kept in citrate buffer overnight at 4°C or immediately processed. An individual flower bud dissected in a drop of 60% acetic acid on a microscope slide under a stereomicroscope was spread on the slide placed on a metal hot plate (50°C) in an additional amount of 60% acetic acid (20 μl in total) for c. 30 s. Then, the preparation was fixed in freshly prepared ethanol : acetic acid (3 : 1) by dropping the fixative around the drop of acetic acid and into it. The preparation was dried using a hair dryer and staged using a phase contrast microscope. Suitable slides were postfixed in freshly prepared 4% formaldehyde in distilled water for 10 min and put onto a rack to air-dry. Preparations were kept in a dust-free box at room temperature until used.
Genomic in situ hybridization was performed on mitotic preparations of C. flexuosa. To remove cytoplasm before GISH and CCP/GISH, the slides were treated with pepsin (0.1 mg ml−1; Sigma) in 0.01 M HCl for 10 min, postfixed in 4% formaldehyde in 2 × SSC (2 × SSC: 3 M sodium chloride, 300 mM trisodium citrate, pH 7.0) for 10 min, and dehydrated in an ethanol series (70, 80 and 96%). For GISH, total genomic DNA (gDNA) was extracted from healthy young leaves of C. amara (Urnerboden), C. hirsuta (Oxford), C. impatiens and C. parviflora according to Dellaporta et al. (1983). Isolated gDNA of the four Cardamine spp. was labelled with either biotin-dUTP or digoxigenin-dUTP via nick translation as described by Mandáková et al. (2010). A quantity (400 ng) of each labelled gDNA probe was ethanol-precipitated and used as GISH probes. The pellet was resuspended in 20 μl of hybridization mix (50% formamide and 10% dextran sulphate in 2 × SSC) per slide; no blocking DNA had been applied. The probe and chromosomes were denatured together on a hot plate at 80°C for 2 min and incubated in a moist chamber at 37°C overnight. Post-hybridization washing was performed in 20% formamide in 2 × SSC at 42°C. Immunodetection of hybridized DNA probes was as follows: biotin-dUTP was detected by avidin–Texas Red (Vector Laboratories, Burlingame, CA, USA) and amplified by goat anti-avidin–biotin (Vector Laboratories) and avidin–Texas Red; digoxigenin-dUTP was detected by mouse antidigoxigenin (Jackson ImmunoResearch, West Grove, PA, USA) and goat anti-mouse–Alexa Fluor 488 (Molecular Probes, Eugene, OR, USA). Chromosomes were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; 2 μg ml−1) in Vectashield (Vector Laboratories). Fluorescence signals were analysed using an Olympus BX-61 epifluorescence microscope and CoolCube CCD camera (MetaSystems, Altlussheim, Germany). Images were acquired separately for the three fluorochromes using appropriate excitation and emission filters (AHF Analysentechnik, Tübingen, Germany). The three monochromatic images were pseudocolored and merged using Adobe Photoshop CS2 software (Adobe Systems, San Jose, CA, USA).
Comparative chromosome painting was performed on C. amara and C. hirsuta pachytene chromosomes. For CCP, on average, each third Arabidopsis thaliana BAC clone was used to establish contigs corresponding to the 24 genomic blocks of the ACK (Schranz et al., 2006). For the detailed composition of the BAC contigs, see Mandáková et al. (2010). The recently reconstructed cytomolecular map of C. amara (Mandáková et al., 2013) was used to navigate design of CCP probes in both C. amara and C. hirsuta accessions analysed herein. The A. thaliana BAC clone T15P10 (AF167571) containing 45S rRNA genes was used for in situ localization of nucleolar organizer regions (NORs), and the A. thaliana clone pCT4.2 (M65137), corresponding to a 500 bp 5S rDNA repeat, was used for localization of 5S rDNA loci. The Arabidopsis-type telomere repeat (TTTAGGG)n was prepared according to Ijdo et al. (1991). All DNA probes were labelled with biotin-dUTP, digoxigenin-dUTP or Cy3-dUTP by nick translation as described by Mandáková et al. (2010). Quantities of 200 ng of each labelled BAC clone and 100 ng of labelled 45S rDNA, 5S rDNA and telomeric repeat were pooled together and precipitated. The subsequent steps are identical to those described earlier. Cy3-dUTP-labelled probes were observed directly.
Comparative chromosome painting /GISH in C. flexuosa included CCP on mitotic and pachytene chromosomes followed by GISH using labelled gDNA of C. amara and C. hirsuta. Painting probes corresponding to 24 ancestral genomic blocks were designed according to the reconstructed karyotypes of C. amara and C. hirsuta. After initial experiments revealed a rearrangement involving genomic block I, the BAC contig spanning this block was split into smaller, differentially labelled subcontigs (Fig. 3). CCP followed the protocol described in the preceding section. After CCP hybridization patterns were analysed and photographed, painting probes were removed by washing the slides in 2 × SSC, and 20% formamide in 2 × SSC at 42°C. Subsequently, the slides were postfixed in 4% formaldehyde in 2 × SSC and dehydrated in an ethanol series (70, 80, and 96%). GISH was performed according to the protocol described. Painted pachytene chromosomes in Fig. 3 were straightened using the plugin ‘Straighten Curved Objects’ (Kocsis et al., 1991) in the Image J software (US National Institutes of Health, Bethesda, MD, USA).
Parentage of C. flexuosa
By chromosomal counting of DAPI-stained mitoses from anther tissues, all C. flexuosa populations were found to be tetraploid (2n = 4x = 32). Only 16 bivalents were repeatedly observed during male meiosis. Based on the literature (Ellis & Jones, 1969; Lihová et al., 2006) and our own observations, four diploid Cardamine species (2n = 2x = 16) were considered as potential parental genomes of the allotetraploid C. flexuosa: C. amara, C. hirsuta, C. impatiens and C. parviflora. Labelled genomic DNA (gDNA) of the four diploids was used in different combinations for GISH experiments. At the given stringency of posthybridization washing, only gDNA of C. amara and C. hirsuta hybridized to the chromosomes of C. flexuosa. The gDNA probe from C. amara labelled 14 and gDNA from C. hirsuta hybridized to 16 mitotic chromosomes, and one chromosome pair was labelled by both genomic probes (Fig. 1). This result suggested the existence of an intergenomic rearrangement that was subsequently analysed in detail by CCP/GISH.
Genomes of C. amara and C. hirsuta
Comparative cytomolecular maps were constructed for the two parental diploid species, C. amara and C. hirsuta (Fig. 2), by multicolour CCP using chromosome-specific A. thaliana BAC contigs that cover the entire chromosomes minus centromeres (i.e. c. 95.1 Mb of the A. thaliana genome; www.tair.org). Differentially labelled BAC contigs were combined to build up the individual eight chromosomes (CA1–CA8) and 24 ancestral genomic blocks (A–X) of the karyotype of C. amara ssp. amara (Urnerboden accession, Mandáková et al., 2013). All Arabidopsis painting probes hybridized reliably to eight chromosomes in both Cardamine species and revealed extensive chromosome collinearity between Cardamine genomes and the ACK (Schranz et al., 2006). Both Cardamine species and ACK shared six chromosomes with conserved ancestral collinearity; only chromosome CA1 differed by a 13 Mb pericentric inversion specific for C. amara. Both parental karyotypes differed from ACK by a whole-arm reciprocal translocation between homeologues of ancestral chromosomes AK6 and AK8. This karyotype was invariable in three different populations analysed in both diploid species. In relation to 95.1 Mb of the A. thaliana genome, the parental genomes exhibited interspecies genome collinearity of c. 86%, with seven out of eight chromosomes being perfectly collinear between C. amara and C. hirsuta (Fig. 2).
Genome structure of C. flexuosa revealed by CCP/GISH
Knowing the structure of C. amara and C. hirsuta chromosomes, corresponding painting probes were hybridized to pachytene chromosomes of C. flexuosa, followed by GISH, addressing the origin of all the chromosomes from both parental species. Out of the 16 chromosome pairs, 14 were structurally identical with chromosomes within the parental genomes (see ideograms in Fig. 2). Two chromosome pairs displayed a different arrangement of genomic blocks as compared with both progenitor genomes. As one arm of both chromosomes comprised genomic block J, these were identified as rearranged chromosomes CA4 and CH4. The structures of both chromosomes, including the organization of genomic blocks, were analysed in detail (Fig. 3). CCP/GISH analysis showed that both rearranged homeologues have undergone an identical pericentric inversion followed by a reciprocal translocation with breakpoints within both genomic blocks I, exchanging unequal proportions of the upper arms of progenitor chromosomes CA4 and CH4. Both chromosomes, referred to as translocation chromosomes t(CA4) and t(CH4) hereafter, shared altered collinearity of blocks I and J, and differed by the length of their upper arms (Figs 1-3). In both chromosomes, an c. 1 Mb segment of block I was found located on the bottom arms, adjacent to the t(CA4) and t(CH4) centromeres. In t(CA4), the upper arm comprised the remaining part of block IA (c. 1.8 Mb) from the original chromosome CA4 and 2 Mb of the IH segment translocated from CH4 (C. hirsuta). A heterochromatic knob containing 5S rDNA repeats was situated between the IA and IH segments. In t(CH4), the upper arm comprised a 0.4 Mb segment of block IA translocated from chromosome CA4, a short (0.1 Mb) IH segment from CH4, and a heterochromatic knob at the terminal position (Figs 2, 3).
Origin of translocation chromosomes t(CA4) and t(CH4)
The origin of two translocation chromosomes in C. flexuosa was inferred from CCP/GISH data (Fig. 3) and is summarized in Fig. 4. The identical structure of both bottom arms implies that the first step included nonhomologous pairing between chromosome CA4 and CH4, and a pericentric inversion with breakpoints within block I and the pericentromere region adjacent to block J. The pericentric inversion was most probably followed by a reciprocal (and intergenomic) translocation with breakpoints within block I on CA4, and proximally adjacent to the 5S rDNA locus on CH4. This is supported by the 5S rDNA locus being located at the junction of IA and IH segments on t(CA4) and in its original position at the CH4 pericentromere in C. hirsuta (Fig. 2). Translocation altered the size and arm ratios between the two acrocentric AK4 homeologues (upper arm, 3.3 Mb; bottom arm, 6.3 Mb). Whereas the t(CA4) chromosome changed its morphology only slightly (upper arm, 3.9 Mb; bottom arm, 7.3 Mb), the arm ratio of t(CH4) changed significantly (upper arm, 0.5 Mb; bottom arm, 7.3 Mb). Eventually, the small 0.5 Mb upper arm of t(CH4) was reversed by a paracentric inversion, yielding a heterochromatic knob at the chromosome terminus.
Intraspecific karyotypic stasis in C. flexuosa
Eight different populations of C. flexuosa from across its European distribution area (Fig. S1) were analysed for intraspecific chromosomal polymorphism using CCP/GISH. Detailed inspection of all 16 chromosome pairs did not reveal any population-specific chromosome rearrangements.
Variation in rDNA loci and genome size
To investigate several heterochromatic chromosomal knobs in the three species studied, 5S and 45S rDNA and an Arabidopsis-type telomeric repeat were localized by FISH. The telomeric probe hybridized only to chromosome ends and no interstitial signals were detected in all three species analysed (data not shown). 45S rDNA repeats comprised the terminal nucleolar organizer regions (NOR) on the short arms of chromosomes CA3 and CA7 in C. amara, and on chromosomes CH2, CH4 and CH6 in C. hirsuta (Fig. 2a). One 5S rDNA locus was localized to the pericentromere regions of chromosome CA8 in C. amara and CH4 in C. hirsuta. In C. flexuosa, CCP/GISH mapping permitted all rDNA loci to be assigned to individual chromosomes. The two 45S rDNA loci in C. flexuosa were inherited from C. amara as terminal NORs on homeologues CA3 and CA7. None of the C. hirsuta subgenome chromosomes carried a 45S rDNA locus. By contrast, the 5S rDNA locus on CA8 was eliminated in C. flexuosa and the locus on CH4 was transposed to chromosome t(CA4) by the intergenomic translocation described earlier. No variations in the number and localization of rDNA repeats were observed among investigated populations of C. amara, C. hirsuta and C. flexuosa.
Complex genomes of polyploid species can be mapped and subgenomes dissected using several complementary approaches. Here we showed that along with genetic mapping and whole-genome sequencing, physical mapping through CCP can be successfully employed to characterize subgenome complements of an allopolyploid plant species. Moreover, combined CCP and GISH analysis on extended pachytene chromosomes allows for detailed analysis of intra- and intergenomic rearrangements, such as unequal reciprocal translocations, which could be erroneously interpreted as nonreciprocal translocations if only GISH is applied.
The origin of the allotetraploid C. flexuosa
The allotetraploid C. flexuosa, a usually biennial (or an overwintering annual) and autogamous species, is distributed throughout the European continent from whence it was introduced to North America. It occurs mostly in natural or seminatural habitats; nevertheless there are also reports of its occurrence in plant nurseries as a weed (Post et al., 2011). Here we have shown that C. flexuosa is an amphidiploid species resulting from hybridization between two diploid species, C. amara and C. hirsuta. Based on chloroplast DNA variation, C. amara is most likely the maternal parent of C. flexuosa (Lihová et al., 2006). C. amara comprises several infraspecific taxa at the diploid and tetraploid levels. Out of these, three diploid subspecies were included in our study, namely C. amara ssp. amara (widespread in Europe), ssp. opicii (restricted to the Carpathians and Sudeten mountains) and ssp. balcanica (occurring in the Balkan mountains of Bulgaria, Serbia, Macedonia and Greece). This perennial and outcrossing species occurs mostly in wetland habitats and is restricted in its distribution to the natural localities. By contrast, C. hirsuta is an annual, autogamous and weedy species, generally occurring in open habitats. As this species is able to occupy new areas very quickly and effectively, currently it has almost cosmopolitan distribution. The original distribution area of C. hirsuta is not exactly known, but was most likely southern or central Europe.
As both parental species and C. flexuosa are widespread in Europe, we argue that the hybridization occurred on the European continent. The identical genome structure, including the translocation chromosomes t(CA4) and t(CH4), revealed in all eight analysed populations, suggests a single origin of C. flexuosa. Although recurrent origins of allopolyploid species are frequent (Soltis & Soltis, 2009), single origins of allopolyploids occur too. A single origin was, for example, suggested for Arabidopsis suecica (Jakobsson et al., 2006, 2007), for two Galeopsis allopolyploids (Bendiksby et al., 2011), for Spartina anglica (Ainouche et al., 2004) and for other allopolyploids.
The divergence age of the tribe Cardamineae and genus Cardamine remains controversial. The age of the genus Cardamine was estimated as c. 6 million yr (Carlsen et al., 2009), but a recent analysis (Beilstein et al., 2010) suggests much earlier divergence dates for Brassicaceae. As the genome structure of Swedish and Irish accessions was identical to that in the remaining European populations, the origin of C. flexuosa can be dated before the end of the last glaciation (c. 10 000 yr ago), when the continental ice shield retreated and Fennoscandia and Ireland were recolonized from southern refugia. The origin of C. flexuosa is, to some extent, comparable with the origin of A. suecica (2n = 4x = 26) – an allopolyploid species resulting from hybridization between A. thaliana (2n = 2x = 10) and Arabidopsis arenosa (2n = 2x, 4x = 16, 32). Based on variations in nuclear and chloroplast markers, A. suecica most probably originated from a single hybridization event somewhere in Europe between 10 000 and 50 000 yr ago (Jakobsson et al., 2006, 2007). Considering the ease of GISH and the strength of hybridization signals (Fig. 1), as well as the structural stasis of both parental subgenomes (Fig. 2), C. flexuosa may be regarded as a relatively young allopolyploid species. By extrapolating time-travelling GISH experiments in Nicotiana (Clarkson et al., 2005; Lim et al., 2007a) and dated molecular phylogenies in Beilstein et al. (2010), the origin of C. flexuosa must be younger than 5 million yr, as this period is the upper limit for a successful identification of parental chromosomes by GISH in Nicotiana allopolyploids (Lim et al., 2007a). Thus, it is likely that the origin of C. flexuosa lies between 10 000 and 1.0 million yr ago.
Subgenome stasis in C. flexuosa
The accumulation of intergenomic rearrangements should vary among allopolyploid species, as these differ by age, genome features, population size, life form, reproductive system, and so on. Without information on the age of C. flexuosa, it is almost impossible to appraise whether one intergenomic translocation since the time of allopolypoidization represents a rare event or an event of an average frequency. However, taking into account the largely conserved chromosome collinearity shared among the parental species and C. flexuosa, both subgenomes exhibit considerable stasis within the allopolyploid genome. This is astonishing, considering the possibility of intergenomic repatterning via homeologous recombination as a result of almost complete interspecies genome collinearity between the parental species (Fig. 2). As diploid-like meiotic pairing was consistently observed in C. flexuosa, either homeologous chromosomes are sufficiently divergent at the sequence level or a genetic control similar to that exerted by the Ph1 regulator in the hexaploid bread wheat exists in C. flexuosa (Cifuentes et al., 2010). Paun et al. (2009) summarized and corroborated the previous idea that a greater genetic divergence between parental species is positively correlated with the formation of allopolyploids (vs homoploids) and regular, quasidiploid chromosome pairing. Indeed, the parental species of C. flexuosa, C. amara and C. hirsuta are phylogenetically rather distant, as shown by analyses of nuclear and chloroplast genes (Lihová et al., 2006; Carlsen et al., 2009) and the efficiency of subgenome differentiation by GISH. This genetic divergence can explain (at least partially) regular meiotic pairing and the absence of more extensive intergenomic reshuffling in C. flexuosa.
The stability of parental subgenomes and, conversely, the occurrence of intra- and intergenomic rearrangements, vary among the analysed allopolyploid species and synthetic lines. The reason(s) for either of these phenomena remain to be uncovered. Whereas an overall stability of parental subgenomes, as observed in C. flexuosa, was also reported in cotton (Liu et al., 2001), S. anglica (Ainouche et al., 2004), young Nicotiana tetraploids (Lim et al., 2004), common wheat (Zhang et al., 2013), Milium montianum (Bennett et al., 1992) and Primula egaliksensis (Guggisberg et al., 2008), extensive chromosome reshuffling and/or aneuploidy were observed in oats (Hayasaki et al., 2000), Brassica napus (Udall et al., 2005; Xiong et al., 2011; Nicolas et al., 2012), Tragopogon neoallopolyploids (Lim et al., 2008; Chester et al., 2012), Poa jemtlandica (Brysting et al., 2000) and ancient Nicotiana tetraploids (Lim et al., 2004, 2007a). Chromosome rearrangements, including translocations and epigenetic alterations, are supposed to reflect the ‘genomic shock’ associated with a de novo merger of two or more genomes, and should mediate diploidization of the allopolyploid genome. Consequently, if parental genomes are sufficiently divergent and chromosome mispairing is negligible (Paun et al., 2009), genome diploidization should be less essential and have a slower pace. This seems to be true for the allotetraploid genome of C. flexuosa.
t(CA4, CH4) intergenomic translocation: recombination between two homeologous genomic blocks
The single intergenomic translocation detected in C. flexuosa is most likely a consequence of rare pairing and recombination between two homeologous chromosomes (CA4 and CH4). In C. amara and C. hirsuta, both chromosomes exhibit perfect collinearity of genomic blocks (Fig. 2). The scenario inferred in Fig. 4 includes, as the first step, a pericentric inversion in a heterologous quadrivalent formed by chromosomes CA4 and CH4. Inversion breakpoints shared by CA4 and CH4 explain why both translocation chromosomes display the same nonancestral position of blocks I and J (Fig. 2). The subsequent reciprocal translocation between the two homeologues exchanged chromosome regions of unequal size. As revealed by GISH (Fig. 1), superficially the resulting translocation chromosome t(CA4) appears to contain two adjacent blocks I, whereas subtelocentric chromosome t(CH4) appears to retain only block J on its long arm. At least one inversion and one translocation breakpoint were located in repeat-rich regions (Fig. 4). One inversion breakpoint occurred within the pericentromeric heterochromatin of CA4 and CH4, and a translocation breakpoint on CH4 was located at the pericentromere, probably within or close to a 5S rDNA locus. The involvement of the 5S rDNA locus would explain its current position between blocks IA and IH on chromosome t(CA4), although a later transposition of an rDNA repeat to the IA/IH breakpoint cannot be ruled out. Similarly, breakpoints of the paracentric inversion involving the short arm of chromosome t(CH4) occurred in subtelomeric and pericentromeric heterochromatin. This inversion resembles paracentric inversion on the short arm of chromosome 4 in A. thaliana (Fransz et al., 2000), both resulting in the origin of an interstitial (A. thaliana) or terminal (C. flexuosa) heterochromatic knob. In summary, inversion and translocation events involving chromosomes CA4 and CH4, and their translocation derivatives, predominantly occurred in heterochromatic repeat-rich regions, including the 5S rDNA site, as observed previously for other species (Raskina et al., 2008; Schubert & Lysak, 2011 and references therein).
The t(CA4, CH4) intergenomic translocation is the onset of genome-wide and long-term diploidization of the allotetraploid C. flexuosa genome. In interspecies hybrids and allopolyploid species, homeologous recombination driven by ancestral sequence homology and/or genetic regulators such as Ph1 mutation may occur at different time points after polyploidization (Kopecký et al., 2010; Xiong et al., 2011; Nicolas et al., 2012). Recombination between homeologous chromosome regions accounting for adjoining blocks IA and IH on t(CA4) also mediates contiguous associations of homeologous genomic regions in other ancient or more recent polyploids (Mandáková et al., 2010; X. Wang et al., 2011b; Cheng et al., 2013).
Genome-specific loss of rDNA repeats
In situ localization of rDNA repeats in C. flexuosa showed that all three 45S rRNA gene loci (terminal NORs) from C. hirsuta were eliminated, whereas both NORs from C. amara were retained on the same chromosomes. By contrast, one interstitial 5S rDNA locus from C. hirsuta was retained and moved to translocation chromosome t(CA4), whereas the 5S rDNA locus from C. amara was lost (Fig. 2). This means that the total number of 5S and 45S rDNA loci in C. flexuosa has returned to the number of rDNA loci in the maternal progenitor genome of C. amara. Uniparental loss of the rDNA loci in allopolyploids is not rare (Kotseruba et al., 2003; Clarkson et al., 2005; Lim et al., 2007a; Guggisberg et al., 2008; Malinska et al., 2010; Mlinarec et al., 2012; Weiss-Schneeweiss et al., 2012). Given that losses of gene copies occur, to some extent, even in synthetic lines (Pontes et al., 2004; Malinska et al., 2010; Ksiazczyk et al., 2011; Xiong et al., 2011), it is difficult to draw any conclusion regarding the time-frame of these changes. However, considering that complete uniparental deletions were never observed in synthetic allopolyploids, one can assume that deletion of NORs is a gradual evolutionary process requiring many generations to complete. In contrast to cotton (Wendel et al., 1995) and tobacco (Kovarik et al., 2004), which homogenized their parental rDNA units by gene conversion without changes in locus number, C. flexuosa seems to have decreased the number of 45S loci, as reported for Zingeria (Kotseruba et al., 2003) and Iris (Lim et al., 2007b) polyploids. Both gene conversion and elimination processes are associated with a reduction in gene copies (Kovarik et al., 2008), restoring the rRNA gene dosage balance in polyploid genomes (Birchler & Veitia, 2012). According to the nucleocytoplasmic interaction (NCI) hypothesis of Gill (1991), in allopolyploid genomes, the paternal nuclear genome should experience more profound restructuring than the maternal nuclear and cytoplasmic genomes. While this may hold true for the 45S locus, it is certainly not the case for the 5S locus that had been eliminated from the maternal genome. In addition, there are numerous examples of more extensive rearrangements affecting the maternal subgenome (Lim et al., 2004; Guggisberg et al., 2008; Mlinarec et al., 2012). Thus, the NCI hypothesis cannot explain biased elimination of rDNA loci, and other, perhaps epigenetic, factors (Kovarik et al., 2008) should be considered.
We thank I. Schubert and A. Kovařík for a critical reading of the manuscript. M. Tullin, M. Tsiantis, K. Mummenhoff and F. Kolář are acknowledged for providing seeds. This work was by supported by research grants from the Czech Science Foundation to M.A.L. and K.M. (P501/10/1014), by the European Social Fund (CZ.1.07/2.3.00/20.0189 and CZ.1.07/2.3.00/30.0037), and the European Regional Development Fund (CZ.1.05/1.1.00/02.0068).