Long-term genome diploidization in allopolyploid Nicotiana section Repandae (Solanaceae)


  • James J. Clarkson,

    Corresponding author
    1. Jodrell Laboratory, Royal Botanic Gasrdens, Kew, Richmond, Surrey, TW9 3DS, UK;
    2. School of Biological Sciences, Queen Mary, University of London, Mile End Road, London, E1 4NS, UK;
      Author for correspondence: James J. Clarkson Tel: +44 208 332 5355 Fax: +44 208 332 5310 Email: j.clarkson@rbgkew.org.uk
    Search for more papers by this author
  • K. Yoong Lim,

    1. School of Biological Sciences, Queen Mary, University of London, Mile End Road, London, E1 4NS, UK;
    Search for more papers by this author
    • *

      Contributed as many data as the first author

  • Ales Kovarik,

    1. Institute of Biophysics, Academy of Sciences of the Czech Republic, CZ-61265 Brno, Czech Republic;
    Search for more papers by this author
  • Mark W. Chase,

    1. Jodrell Laboratory, Royal Botanic Gasrdens, Kew, Richmond, Surrey, TW9 3DS, UK;
    Search for more papers by this author
  • Sandra Knapp,

    1. Department of Botany, The Natural History Museum, Cromwell Road, Lsndon, SW7 5BD, UK.
    Search for more papers by this author
  • Andrew R. Leitch

    1. School of Biological Sciences, Queen Mary, University of London, Mile End Road, London, E1 4NS, UK;
    Search for more papers by this author

Author for correspondence: James J. Clarkson Tel: +44 208 332 5355 Fax: +44 208 332 5310 Email: j.clarkson@rbgkew.org.uk


  • • Here, we analyze long-term evolution in Nicotiana allopolyploid section Repandae (the closest living diploids are N. sylvestris, the maternal parent, and N. obtusifolia, the paternal parent). We compare data with other more recently formed Nicotiana allopolyploids.
  • • We investigated 35S and 5S nuclear ribosomal DNA (rDNA) chromosomal location and unit divergence. A molecular clock was applied to the Nicotiana phylogenetic tree to determine allopolyploid ages.
  • • N. tabacum and species of Repandae were c. 0.2 and 4.5 Myr old, respectively. In all Repandae species, the numbers of both 35S and 5S rDNA loci were less than the sum of those of the diploid progenitors. Trees based on 5S rDNA spacer sequences indicated units of only the paternal parent.
  • • In recent Nicotiana allopolyploids, the numbers of rDNA loci equal the sum of those of their progenitors. In the Repandae genomes, diploidization is associated with locus loss. Sequence analysis indicates that 35S and 5S units most closely resemble maternal and paternal progenitors, respectively. In Nicotiana, 4.5 Myr of allopolyploid evolution renders genomic in situ hybridization (GISH) unsuitable for the complete resolution of parental genomes.


The evolution of many, perhaps all, angiosperms, has involved polyploidy. Sequence data have indicated that even some apparently ‘diploid’ species are palaeopolyploids (Lagercrantz & Lydiate, 1996), for example, the ‘diploid’Arabidopsis thaliana (L.) Heynh., with its small genome and only five chromosomes, may be a palaeohexaploid (Vision et al., 2000; Bowers et al., 2003). Realization of the importance of polyploidy in angiosperm evolution has stimulated considerable interest in the genetic consequences of this phenomenon, both in natural and in newly synthesized polyploids. Subsequent research has led to conflicting results, such as the rapid divergence of amplified fragment length polymorphisms (AFLPs) in newly synthesized Triticum allopolyploids, but not in Gossypium allopolyploids. In natural allopolyploids, there is evidence of sequence homogenization of 35S rDNA in Gossypium (Malvaceae), Nicotiana (Solanaceae) and Glycine (Fabaceae), but not in Brassica and Arabidopsis[both Brassicaceae, reviewed in Leitch & Bennett (1997), Liu & Wendel (2003) and Soltis et al. (2004a)].

One reason for the lack of consensus is that the genomic consequences of changes in ploidy may vary as a result of the different ages of the polyploids being studied, which may obscure changes that would otherwise be commonly associated with polyploidy. Leitch & Bennett (2004) observed that ‘genome downsizing’ is a general long-term response to polyploidy in many plant groups, whereby a polyploid genome may, over time, become ever more ‘diploid-like’ in character. The long-term genetic consequences of allopolyploidy and processes by which genomes become ‘diploidized’ are little understood and studied. Here we address these topics by using the polyploid Nicotiana section Repandae (Solanaceae). We take advantage of well-supported phylogenetic trees of the genus (Chase et al., 2003; Clarkson et al., 2004) to predict the closest living diploid progenitors of Repandae and ages of Nicotiana polyploids and sections. We compare our understanding of long-term genome evolution since Repandae origin with shorter-term evolutionary outcomes previously reported for allopolyploid N. tabacum (cultivated tobacco 2n = 4x = 48).

Section Repandae consists of four allopolyploid species –N. nudicaulis, N. repanda, N. stocktonii and N. nesophila (Knapp et al., 2004); all are 2n = 4x = 48. Goodspeed (1954) recognized in Repandae only the last three, with the morphologically divergent N. nudicaulis (see Fig. 3A,B in Knapp et al., 2004) in a monotypic section Nudicaules. Phylogenetic analyses based on internal transcribed spacer (ITS) sequences of nuclear ribosomal DNA (rDNA) (Chase et al., 2003) and plastid sequences (Clarkson et al., 2004) clearly show that Goodspeed's (1954) monotypic section Nudicaules is closely related to Repandae and descended from a unique common ancestor shared with the three species of Repandae. DNA sequence analysis also indicated that an ancestor of N. sylvestris (the sole member of section Sylvestres) was the maternal genome donor of Repandae (Clarkson et al., 2004) and that an ancestor of section Trigonophyllae was the other parent (Clarkson et al., unpublished). Section Trigonophyllae consists of a single species, N. obtusifolia (previously known as N. trigonophylla; Knapp et al., 2004). A second species in this section, N. palmeri, has also been described, but it is so closely related to N. obtusifolia that there were no differences detected in 4656 base pairs (bp) of plastid DNA (Clarkson et al., 2004); the type of N. palmeri falls within the range of variation of N. obtusifolia and has previously been synonymized with it (Wells, 1960). In any case, it would have been a shared common ancestor of both of these that was the parent of section Repandae. Therefore, for practical reasons we use, in the present study, N. sylvestris and N. obtusifolia as the diploid species most closely related to section Repandae.

Figure 3.

Root-tip metaphases after fluorescence in situ hybridization (FISH) with 5S (red) and 35S rDNA (green) probes and chromosome counter-stain with 4′,6-diamidino-2-phenylindole (DAPI). (a) Nicotiana sylvestris; (b) N. obtusifolia; (c) N. nudicaulis; (d) N. repanda; (e) N. nesophila; and (f) N. stocktonii. Genomic in situ hybridization (GISH) to N. nesophila (g) and N. nudicaulis (h), probed with total genomic DNA of N. sylvestris (green) and total genomic DNA of N. obtusifolia (red), and in (g) DAPI labeling (blue) overlaid. In (g) GISH reveals a strong interstitial band with N. obtusifolia genomic DNA probe (arrowhead). Bar, 10 µm.

Section Repandae has a disjunct natural distribution, with the greatest species diversity occurring in Mexico. Nicotiana repanda is the most widespread species (it is found in Texas, Mexico and Cuba). Nicotiana nudicaulis is found in north-eastern Mexico. The remaining two species (N. stocktonii and N. nesophila) are closely related and are found in the Revillagigedo Islands. These three Pacific islands (San Benedicto, Socorro and Clarion) are of volcanic origin from the deep sea Mathematician's Ridge on the Pacific Plate (Allan & Nelson, 1991) and have their origins in the Clarion Fracture Zone. They lie far from the coast, approx. 300 km west of Colima, Mexico, and 350 km south-south-west of Baja California, Mexico, and are still volcanically active. They have been called the Galápagos of Mexico because of their volcanic origin, isolation and endemic faunal and floral elements (Brattstrom, 1990). Phylogenetic analyses revealed that Repandae have their origin in a single hybridization event, subsequently followed by a radiation leading to the four extant species in the section (Clarkson et al., 2004). Nicotiana nudicaulis is sister to the rest of the section and highly genetically as well as morphologically divergent from the other three, which share a much more recent common ancestor (Chase et al., 2003; Clarkson et al., 2004).

Allopolyploids in Nicotiana range from recently formed (including artificially created crosses) to relatively ancient, natural species (Goodspeed, 1954). We aim to determine the ages of allopolyploids within the genus based on their sequence divergence from extant relatives of their progenitor diploids. Conflict exists in the literature about the age of the genus Nicotiana. Uchiyama et al. (1977) estimated that the genus originated 75–100 Myr ago, but the deeper split between Nicotiana and the genus Petunia (also Solanaceae) has subsequently been dated at only 23–25 Myr (Wikström et al., 2001).

Within Nicotiana, two potential dating points can be used to calibrate our molecular phylogenetic trees. Both rely on published dates at which a particular volcanic island rose above sea level. This type of calibration point puts a maximum age on species (Richardson et al., 2001) because speciation of an endemic species could have taken place after the island formed. Nicotiana cordifolia is endemic to the Juan Fernández Islands (or Robinson Crusoe Islands), Chile, and known only from the island of Masafuera. The two principal islands of the archipelago are Masatierra, which is approx. 3.8–4.2 Myr old and closest to the Chilean mainland, and Masafuera, which is approx. 1–2.4 Myr old (Stuessy et al., 1984; Sang et al., 1995). Divergence of N. cordifolia from its mainland (Chilean) sister taxon N. solanifolia can therefore be dated. Nicotiana stocktonii and N. nesophila are endemic to the Revillagigedo Islands, which are between recent (San Benedicto, which was devastated by an eruption in 1952, see Sanchez Rubio, 1981) and approx. 1.1 Myr old (Clarion) or slightly older (Duncan & Clague, 1985). Their divergence from the mainland (Mexican) sister taxon N. repanda can therefore also be dated.

A limited number of Nicotiana species (primarily diploid, 10 in total) have previously been compared, in a phylogenetic context using nontranscribed spacer sequences (NTS) of 5S rDNA (Kitamura et al., 2001), but no species of Repandae were included in that analysis. Here we analyze these sequences from the species of section Repandae. We also use Southern hybridization and fluorescence in situ hybridization (FISH) to determine the organization and distribution of 5S and 35S rDNA in Repandae and the modern diploid relatives of their progenitor lineages.

These data are also compared with patterns of divergence in more recent allopolyploid species, N. tabacum, N. rustica and N. arentsii, which, like Repandae, are all 2n = 4x = 48 [reviewed in Kovarik et al. (2004) and in Lim et al. (2004)]. These three more recently formed allopolyploids (based on their lower levels of sequence divergence in analyses of plastid DNA data) all show some or complete sequence homogenization of 35S rDNA sequences, and each has 5S and 35S rDNA loci numbers that are equal to the sum of their progenitor diploids. Nicotiana tabacum is formed from a cross involving something close to N. sylvestris, the maternal S-genome donor (like Repandae) and N. tomentosiformis, the T-genome donor (Goodspeed, 1954; Kitamura et al., 2000; Chase et al., 2003; Murad et al., 2005). Genomic in situ hybridization (GISH) clearly identifies the parental S and T genomes (Parokonny et al., 1992; Kenton et al., 1993; Lim et al., 2000b). In synthetic tobacco, concerted evolution of 35S rDNA occurs rapidly and involves thousands of units that become altered in only four generations (Skalicka et al., 2003). These young synthetic tobacco plants also show evidence for preferential loss of repetitive sequences from the paternal genome (Skalicka et al., 2005). A similar phenomenon was previously reported from restriction fragment length polymorphisms (RFLPs) in synthetic Brassica allotetraploids. In Brassica, the preferential loss of RFLPs of paternal origin was ascribed to adverse interactions of the paternal genome in a maternal cytoplasm (the nuclear cytoplasmic interaction hypothesis). We compare some of these data with the long-term consequences of allopolyploidy on genome evolution in Repandae.

Materials and Methods

DNA sequencing

For DNA extractions, 1 g of fresh tissue or 0.3 g of silica-gel dried material was used. All accessions are vouchered at the Natural History Museum, London (BM; Table 1).

Table 1.  Voucher and GenBank numbers for NTS sequences are given for all species of Nicotiana studied
SpeciesVoucher detailsGenBank accession
  • a

    Sequence that was downloaded from GenBank.

Nicotiana alata Link & OttoHelgason & Monro 501, 518 BMAJ849919
Nicotiana attenuata Torrey ex S. Wats.Helgason & Monro 621 BMAJ849923
Nicotiana bonariensis Lehm.Helgason & Monro 622 BMAJ849920
Nicotiana cordifolia Phil.Saikia 008 BMAJ849927
Nicotiana linearis Phil.Helgason & Monro 609 BMAJ849925
Nicotiana miersii J. RémyClarkson 003 BMAJ849924
Nicotiana nesophila I.M. Johnst.Saikia 011 BMAJ849928
Nicotiana noctiflora Hook.Lim 005 BMAJ849921
Nicotiana nudicaulis S. Wats.Saikia 004 BMAJ849931
Nicotiana obtusifolia M. Martens & GaleottiLim 004 BMAJ849933
Nicotiana otophora Griseb.AJ131172a
Nicotiana palmeri A. GrayHelgason & Monro 631 BMAJ849932
Nicotiana petunioides (Griseb.) MillánLim 001 BMAJ849922
Nicotiana repanda Willd.Clarkson 019 BMAJ849930
Nicotiana stocktonii BrandegeeLim 003 BMAJ849929
Nicotiana sylvestris Speg. & ComesHelgason & Monro 628 BMAJ849918
Nicotiana tomentosiformis Goodsp.AJ131179a
Nicotiana wigandioides Koch & Fintelm.Nee et al. 51764 BMAJ849926

DNA was extracted by using a modified 2× cetyltrimethylammonium bromide (CTAB) method (Doyle & Doyle, 1987). DNA was precipitated in chilled ethanol (−20°C) for at least 24 h and then resuspended in 1.55 g ml−1 cesium chloride/ethidium bromide. Samples were then purified using a density gradient, followed by removal of the ethidium bromide and cesium chloride with butanol/dialysis and then storage in Tris–EDTA (TE).

Target regions were amplified in a Gene Amp™ 9700 polymerase chain reaction (PCR) system (ABI; Applied Biosystems, Inc., Warrington, Cheshire, UK), using ReddyMix PCR Mastermix, at 2.5 mm MgCl2 (ABgene, Epsom, Surrey, UK). Details of primers were as described previously (Matyasek et al., 2002). The PCR program consisted of 4 min at 94°C followed by 28 cycles of 1 min at 94°C, 1 min at 50°C, and 1 min at 72°C. Dimethylsulphoxide (DMSO) (5%, v/v) was used in both PCR and cycle sequencing reactions to remove the effects of secondary structure caused by a high GC content, which makes primer annealing and sequencing difficult (see Chase et al., 2003 for more details). PCR products were cleaned on Nucleospin® miniprep columns (Macherey–Nagel, Duren, Germany) following the manufacturer's protocols. Samples were sequenced on an ABI 3100 capillary DNA sequencer using Big Dye terminator v3.1 chemistry, following the manufacturer's protocols (ABI). For cleaning, we used precipitation in ethanol, etc.

The raw sequences were edited and assembled using sequencher version 4.1 (Gene Codes, Inc., Ann Arbor, MI, USA). They were aligned by eye following the guidelines provided by Kelchner (2000). Gaps were coded as missing data; they were thus excluded from the analysis. We did, however, survey their distribution after the analysis to determine whether their inclusion would provide improved resolution or bootstrap support, but in no case were they useful in this way; they routinely marked clades for which bootstrap support was already high, based simply on analyses of nucleotide substitutions alone.

Parsimony analysis

paup, version 4.0b (Swofford, 2001), was used for parsimony analysis. Eighteen accessions were included in searches (16 of them new in this study), and 180 bp had to be excluded from the sequences as a result of ambiguous alignment. Branch-and-bound searches were performed, and all character transformations were treated as equally likely and unordered (Fitch, 1971). DELTRAN character optimization was used to illustrate branch lengths throughout (owing to reported errors with ACCTRAN optimization in paup version 4.0b). Nicotiana tomentosiformis and N. otophora (representatives of section Tomentosae) were designated as outgroups. Section Tomentosae has consistently been shown to be a sister to the rest of Nicotiana (Chase et al., 2003; Clarkson et al., 2004). These two sequences were downloaded from GenBank (see Table 1 for accession numbers). To assess internal support, 100 bootstrap replicates (Felsenstein, 1985) were performed by using the branch-and-bound algorithm with equal weights. We follow the new sectional classification of Nicotiana throughout (Knapp et al., 2004), rather than that of Goodspeed (1954).

Molecular clock methods

Our previously published combined plastid and ITS data set (Clarkson et al., 2004) was used for the dating exercise because this is the most robust analysis obtained, to date, for the genus (judged by bootstrap percentages). This data set encompassed only diploid taxa and so we had to add the four allopolyploids of Repandae and N. tabacum. Both data sets are congruent for the phylogenetic position of the Repandae species because the ITS region has been converted in all species to the maternal type (the same parent as the plastid donor). However, for N. tabacum, the ITS sequence had to be excluded because the two data sets are incongruent. Nicotiana tabacum inherited its plastid genome from its maternal parent (N. sylvestris), and the ITS has been converted to that of its paternal parent (N. tomentosiformis; Lim et al., 2000a; Chase et al., 2003). As N. tabacum and N. tomentosiformis have identical ITS sequences, the removal of these data will not affect the overall estimation of its time of origin. After the addition of the five extra allopolyploids to the combined data set, analysis produced just a single most-parsimonious tree (the same number as yielded by the original published analysis). The relative rates test (Felsenstein, 1981) showed that divergence was not proportional to time. Therefore, nonparametric rate-smoothing (NPRS) (Sanderson, 1997) was used to smooth out rate heterogeneity within this tree. Trees were calibrated using 2.4 Myr for the split between N. solanifolia and N. cordifolia because this is the more divergent, and therefore presumably the more reliable, of the two potential calibration points (see Richardson et al., 2001). Scaling the tree according to the longer of the two available branches reduces the effect of errors. By using this approach, one calibration point can be used to see how well it corroborates the other.

Plant material and chromosome preparations

Root tips of N. repanda, N. stocktonii, N. nudicaulis, N. nesophila, N. sylvestris and N. obtusifolia (Table 1) were used for FISH. Root tips from pot-grown plants were analyzed following pretreatment with a saturated aqueous solution of Gammexane® (hexachlorocyclohexane; Sigma Aldrich, Poole, UK) for 4 h. All root-tips were fixed in absolute ethanol : glacial acetic acid (3 : 1, v/v) for longer than 1 wk. Chromosome squashes were prepared following the enzymatic softening of material, as described previously (Leitch et al., 2001).

Probes for FISH

  • 1Nicotiana sylvestris total genomic DNA was labeled by nick translation with digoxigenin-11-dUTP (Roche Biochemicals, Lewes, UK), and N. obtusifolia total genomic DNA was labeled by nick translation with biotin-16-dUTP (Sigma Aldrich), according to Leitch et al. (2001).
  • 2The entire plasmid containing pTa71 [isolated by Gerlach & Bedbrook (1979)], which includes the 18-5.8-26S rDNA subunits and the intergenic spacer isolated from Triticum aestivum, was labeled with digoxigenin-11-dUTP by nick translation.
  • 3The coding portion of the 5S region (approx. 120 bp) was amplified from N. rustica by following the PCR program and primers described by Fulnecek et al. (1998). Biotin-16-dUTP was used to PCR label the sequence, according to the methods of Schwarzacher & Heslop-Harrison (2000), by using ×10 concentrations of dNTPS and dUTP. Nicotiana sylvestris was probed with pTZ-19R [5S rDNA; Venkateswarlu et al. (1991)].


FISH was carried out as described in Leitch et al. (1994), with modifications as described in Lim et al. (1998). Briefly, slides were pretreated with 100 µg ml−1 RNase A for 1 h and then with 0.25 µg ml−1 pepsin for 5 min, followed by denaturation in 70% (v/v) formamide in ×2 SSC (0.3 m sodium chloride, 0.03 m sodium citrate), at 70°C for 2 min. The hybridization mix contained 50% (v/v) formamide, 10% (w/v) dextran sulphate and 0.1% (w/v) sodium dodecyl sulphate in ×2 SSC. For GISH, the hybridization mixture contained 8 µg ml−1 digoxigenin-labeled N. sylvestris DNA and 8 µg ml−1 biotin-labelled N. obtusifolia DNA. A concentration of 4 µg ml−1 DNA was used in the mixture when using digoxigenin-labelled pTa71 and biotin-labelled 5S rDNA PCR product (or pTZ19-R). After overnight hybridization at 37°C, slides were washed in 20% (v/v) formamide in 0.1 × SSC at 42°C at an estimated hybridization stringency of 80–85%. Sites of probe hybridization were detected by using 20 µg ml−1 fluorescein-conjugated anti-digoxigenin IgG (Roche Biochemicals) and 5 µg ml−1 Cy3-conjugated avidin (Amersham Biosciences, Little Chalfont, UK). Chromosomes were counterstained with 2 µg ml−1 4′,6-diamidino-2-phenylindole (DAPI) in 4 × SSC, mounted in Vectashield medium (Vector Laboratories, Peterborough, UK), examined using a Leica DMRA2 epifluorescence microscope (Wetzlar, Germany) and photographed with an Orca ER camera (Hamamatsu Photonics, Welwyn Garden City, UK). Images were processed for colour balance, contrast and brightness uniformity.

Restriction endonuclease digestion and Southern hybridization

Genomic DNA was extracted from the young leaves of these plants by using the method of Saghai-Maroof et al. (1984) with modifications according to Kovarik et al. (1996). The methods used were those described in Sambrook et al. (1989). DNA samples were digested to completion with excess restriction endonuclease and fractionated in 1% (w/v) agarose by gel electrophoresis. DNA was transferred to a Hybond N+ membrane (Amersham Pharmacia Biotech., Little Chalfont, UK). Southern hybridization was carried out under high-stringency conditions (Fulnecek et al., 2002) by using a heat-denatured 32P-labeled rDNA probe that was either (i) a 1.7-kb EcoRI fragment of the 18S rRNA gene subunit from Solanum lycopersicum L. (tomato, accession number X51576; Kiss et al., 1989a) or (ii) a 220-bp fragment of the 26S rRNA gene subunit from S. lycopersicum (accession number X13557; Kiss et al., 1989b) obtained by PCR amplification of the region between +2901 (5′-GAATTCACCCAAGTGTTGGGAT-3′) and +3121 (5′-AGAGGCGTTCAGTCATAATC-3′) with respect to the start of the 26S rDNA subunit (Kiss et al., 1989b). Labeling of DNA probes was carried out by a random primer method using [32P]dCTP (Dekaprime™ kit; Fermentas, Vilnius, Lithuania).


NTS sequences

The NTS region of 5S rDNA has a high frequency of informative positions, which more than makes up for its short length (358 bp included in analysis), producing well-resolved trees (Fig. 1). The total number of characters was 358, of which 188 (53%) were variable and 128 (36%) were potentially parsimony informative. The percentage of parsimony-informative characters (36%) is favorable when compared to other data sets for Nicotiana, for example 27% for ITS (Chase et al., 2003) and only 8.8% for the plastid data set (Clarkson et al., 2004). Analysis produced one most-parsimonious tree [length 277 steps; consistency index (CI) 0.87; retention index (RI) 0.88]. The tree presented in Fig. 1 has branch lengths above (DELTRAN optimization) and bootstrap percentages (BP) below the branches.

Figure 1.

The single most-parsimonious nontranscribed spacer (NTS) tree. Branch lengths (DELTRAN optimization) are shown above branches, and all bootstrap percentages (BP) of > 50 are shown below branches. Open bars, diploid sections; closed bar, allopolyploid section Repandae[the sectional classification of Knapp et al. (2004) is used].

Sections Tomentosae (BP 100), Trigonophyllae (BP 100), Repandae (BP 67), Noctiflorae (BP 100), Alatae (BP 100), and Petunioides (BP 89) all form clades. Sections Paniculatae and Undulatae together form a moderately supported clade (BP 85). Section Sylvestres is sister to section Noctiflorae (BP 86). Section Tomentosae was specified as the operational outgroup, based on other evidence (Chase et al., 2003; Clarkson et al., 2004).

Molecular clock

The node that separates N. stocktonii and N. nesophila from N. repanda (second potential calibration point – see Fig. 2) was dated at 1.0 Myr using the rate-smoothed molecular data set. This is consistent with the geological data that predicted it to be approx. 1.1 Myr. The divergence time between Repandae and its maternal progenitor (N. sylvestris) is dated at 4.5 Myr. Nicotiana tabacum originated within the last 0.2 Myr. The genera Nicotiana and Symonanthus (tribe Anthocercideae) are sister groups (Clarkson et al., 2004) that split approx. 15.3 Myr ago (data not shown). Dates from the rate-smoothed tree not pertaining to the discussion of evolution in Repandae will be dealt with elsewhere (J. Clarkson et al., in preparation).

Figure 2.

The ‘combined tree’, from Clarkson et al. (2004), after nonparametric rate smoothing (NPRS). Branch lengths are proportional to time, with scale in millions of years. Note that: only the relevant clade is presented.

35S and 5S rDNA chromosome distribution

FISH analyses of mitotic metaphases of N. sylvestris reveal three subtelomeric rDNA loci on the short arms of chromosomes 10, 11 and 12, and one 5S rDNA locus interstitial to the long arm of chromosome 8 (Fig. 3a; Lim et al., 2000b). There is a similar distribution of signal in N. obtusifolia (2n = 24), although the 5S DNA locus has an interstitial position on a short-armed chromosome (Fig. 3b). We might therefore have predicted that all members of section Repandae (2n = 4x = 48) would have all of these FISH signals, but N. nudicaulis (Fig. 3c), N. nesophila (Fig. 3e) and N. stocktonii (Fig. 3f) have two and N. repanda (Fig. 3d) has three 35S rDNA loci subtelomeric to the short arm, a situation that resembles many of the diploid species within the genus. Likewise, on a separate chromosome, all four Repandae species have only a single 5S rDNA locus on the short arm, although its precise chromosomal location varies among these species (Fig. 3).

Southern blot hybridization

Southern hybridization to BstNI-restricted DNA revealed that all section Repandae species had one predominant 5S rDNA gene family (Fig. 4a). Three of the species of the group (N. stocktonii, N. repanda, and N. nudicaulis) and N. sylvestris have an indication of a minor band that may represent a low-abundance family of 5S rDNA units. Sizes of the predominant 5S rDNA family were 430 bp for N. sylvestris, as reported by Fulnecek et al. (2002), 600 bp for section Trigonophyllae and 500 bp for all species in section Repandae.

Figure 4.

Southern hybridization of BstNI-restricted genomic DNA probed with (a) 5S rDNA and (b) 26S rDNA.

Southern hybridization analysis of BstNI-restricted DNA, by using the 26S rDNA probe, a method that predominantly reveals polymorphism in the intergenic spacer (IGS), demonstrates that there are three families of 35S rDNA units in N. sylvestris with bands of 2.4, 2.7 and 3.0 kb (Fig. 4b), as reported previously (Lim et al., 2000a). Similarly, restricted DNA of N. obtusifolia (and plants identified as N. palmeri) generates two or three bands that represent distinctive families of 35S rDNA units. No bands in section Repandae species precisely match those from the extant descendents of the progenitor diploids (Fig. 4b).


GISH was conducted on all four species in section Repandae by using labeled N. sylvestris and N. obtusifolia genomic DNAs as probes. In N. repanda, N. stocktonii and N. nesophila, there was no discrimination of the parental origin of chromosomes, although there were scattered dots with the N. sylvestris probe and an interstitial band on a large metacentric chromosome with the N. obtusifolia probe (shown for N. nesophila in Fig. 3g). However, there was some discrimination in N. nudicaulis (Fig. 3h). We attempted to construct a karyotype of GISH-labelled N. nudicaulis metaphase chromosomes in order to differentiate chromosomes of potential N. sylvestris origin and of N. obtusifolia origin. This proved impossible for at least 10 of the 48 chromosomes of the metaphase set owing to blending of the signal. Nevertheless, the majority of the chromosomes did appear to have distinctive genome-specific labeling.


The evolution of polyploid genomes

There is considerable interest in the genetic consequences of polyploidy on gene expression, genome evolution and speciation. Research, to date, has concentrated on three aspects: (1) the immediate consequences of allopolyploidy, with work derived from synthetic allopolyploids resembling natural species (Skalicka et al., 2003; Adams et al., 2004; Joly et al., 2004; Levy & Feldman, 2004; Lukens et al., 2004; Wang et al., 2004); (2) the role of polyploidy in the establishment and evolution of young (less than 150-yr-old) species (Abbott & Lowe, 2004; Ainouche et al., 2004; Soltis et al., 2004b); and (3) the comparison of established polyploid species (thousands or millions of years old) in relation to the closest living diploids (cf. Leitch & Bennett, 1997; Liu & Wendel, 2003; Osborn et al., 2003; Levy & Feldman, 2004; Soltis et al., 2004a). Here we examine a fourth aspect – the divergence of polyploid taxa long after an ancient (c. 4.5 Myr ago) polyploid event.

Finding genetic events that are a unifying response to de novo allopolyploidy is proving difficult, and possibly no event is ubiquitous to all cases, but the existing data point to a number of important features: (1) genetic change in polyploids can occur rapidly (within one or a few generations); (2) changes can influence a large range of sequences (from genes to repeats), but no type of sequence can be expected to change; (3) patterns of change are usually variable between sibling lines; and (4) changes can appear stochastic (e.g. Arabidopsis suecica) or targeted (e.g. N. tabacum, T. aestivum).

In established polyploid species (thousands of years old), genetic change generated by polyploidy is almost certainly fixed, and genetic patterns observed are a combination of this fixed change and subsequent genome divergence. Distinguishing between the two is difficult and requires knowledge of phylogenetic divergence in the diploids and polyploids alike. We aimed to be able to distinguish allopolyploid-induced change from genome divergence by taking advantage of ancient polyploidy and speciation in Repandae. Those genetic patterns that are in common between species represent features of the genome in the common polyploid ancestor. Those features that are different reflect divergence since allopolyploid fixation.

Phylogenetic relationships between diploid Nicotiana species

The Nicotiana species chosen in this analysis were selected to represent at least one species from each of the diploid sections (sensuKnapp et al., 2004). In addition, all species in allopolyploid section Repandae were analyzed. In previous analyses, it has been difficult to establish the relationship between N. sylvestris (the sole member of section Sylvestres) and other diploids owing to a lack of informative variation in the DNA regions analyzed. The most robust ‘diploids-only’ phylogenetic analysis to date, which combined both plastid and ITS sequences (Clarkson et al., 2004), weakly supported N. sylvestris as sister to Alatae (BP 72). However, N. sylvestris has the same chromosome number as section Noctiflorae (n = 12), not Alatae (n = 9,10). Our NTS data resolve N. sylvestris as sister to section Noctiflorae (BP 86). Sections Noctiflorae, Alatae and Petunioides all form a well-supported clade (BP 85), as in the combined (‘diploids only’) analysis (Clarkson et al., 2004). However the inter-relationships between the three sections are different here, with sections Petunioides and Alatae being sister taxa (see Fig. 1), rather than Noctiflorae and Alatae (see Fig. 3 of Clarkson et al., 2004).

Node ages in Nicotiana divergence

After rate smoothing, our data set is internally consistent: one independent calibration point corroborates the other. The age of the split between Nicotiana and Symonanthus (15.3 Myr ago) is in conflict with the results of Uchiyama et al. (1977), who estimated the age of the genus Nicotiana to be 75–100 Myr, based on protein composition analysis. However, these data are, in general, consistent with the more recent findings of Wikström et al. (2001), who dated the split between Nicotiana and Petunia at 23–25 Myr by using DNA sequence data calibrated with one fossil calibration point. These two dates fit together well because Anthocercideae are sister to Nicotiana, and Petunia is a yet more-distant relative (Olmstead et al., 1999). These data show that the hybridization event leading to the formation of N. tabacum occurred less than 0.2 Myr ago. Okamuro & Goldberg (1985) estimated that N. tabacum was 6 Myr old, based on differences between stretches of single-copy DNA in N. tabacum and its two parents. This date seems an extreme overestimation of the age of the species, in view of the fact that we observed no differences between the ITS sequence of N. tabacum and its paternal progenitor (Chase et al., 2003) and only one substitution between N. tabacum and its maternal progenitor in over 4 kb of plastid DNA (Clarkson et al., 2004). In cases in which single-copy nuclear genes have been sequenced for N. tabacum and its progenitor diploids, copies corresponding almost exactly to both progenitors are found in N. tabacum, supporting its recent formation [e.g. putrescine N-methyltransferase (Riechers & Timko, 1999) and heterotrimeric GTP-binding protein (Takumi et al., 2002)]. In addition to this, the chromosome characteristics of N. tabacum are still physically similar to those of its two progenitor diploids, including repeat sequences and integrated viral elements (Lim et al., 2000b; Murad et al., 2002).

Genome evolution in Repandae

Glutamine synthetase (J. Clarkson et al., unpublished) and plastid DNA sequences (Clarkson et al., 2004) have revealed that the closest living relatives of the species of section Repandae are N. obtusifolia (section Trigonophyllae) and N. sylvestris (section Sylvestres), the latter being the maternal genome donor.

GISH labeling of N. nudicaulis with N. sylvestris and N. obtusifolia total genomic DNA revealed some discrimination in label distribution, but entire genome sets could not be readily distinguished. Nevertheless, it was clear that there is discrimination as to genome origin for some chromosomes (Fig. 3). In contrast, no other species in Repandae revealed discrimination between parental chromosome sets and only hybridize weakly across all chromosomes with the N. sylvestris genomic DNA probe and at a large interstitial locus with the N. nudicaulis probe (see Fig. 3). The likely maximum age of section Repandae can be calculated by using the calibrated node on the rate-smoothed tree that separates the mainland N. repanda from its island relatives. We estimate the age of section Repandae to be approx. 4.5 Myr (see Fig. 2). Our results show that these species are old enough to be beyond the point at which GISH is useful in Nicotiana as a method to distinguish parental genomes of allopolyploids.

In more recently formed allopolyploids of Nicotiana (e.g. N. tabacum, N. rustica and N. arentsii), GISH is a highly effective way to discriminate parental genomes (reviewed in Lim et al., 2004). Loss of GISH discrimination over 4.5 Myr of Repandae evolution has presumably arisen as a consequence of diploidization processes that might include genome downsizing (Leitch & Bennett, 2004), sequence mobility (e.g. retroelement, transposon mobility) and sequence divergence, in both allopolyploids and extant diploid species derived from progenitor lineages. All allopolyploid species of Repandae have a common origin, but they differ in the effectiveness of GISH, with N. nudicaulis retaining the greatest overall similarity to progenitor diploids. This differential response among the species of section Repandae indicates that divergence in the allopolyploids has played more of a role in the efficacy of GISH than divergence of the diploid progenitor lineages. At least in Nicotiana, 4.5 Myr appears to be beyond the point at which clear resolution is possible.

5S rDNA evolution in Repandae

Direct NTS sequencing from PCR products confirmed that sections Repandae (BP 67) and Trigonophyllae (BP 100) are sister groups (BP 96; Fig. 1). However, direct sequencing in allopolyploids cannot detect rare classes of rDNA types if present at less than 5% of the predominant type (Rauscher et al., 2002). To confirm that we had good representation of the whole NTS population, we cloned and sequenced a further 10 NTS units per Repandae species (data not shown). All sequences showed the highest similarity to N. obtusifolia, the closest extant relative to the paternal genome parent. In N. sylvestris, the 5S rDNA locus was localized to an interstitial region on the long arm of a small metacentric chromosome. In N. obtusifolia, the locus occurs in subtelomeric positions on the short arm of a metacentric chromosome. In Repandae, the 5S locus occurs in subtelomeric positions, consistent with their N. obtusifolia origin (see Fig. 3 and Table 2). It is likely that one locus of maternal genome origin has been in lost in the species of Repandae. Perhaps reduction in locus number and loss of N. sylvestris-like units is associated with genome diploidization processes.

Table 2.  The numbers of 35S and 5S loci elucidated by fluorescence in situ hybridization (FISH) are summarized for each taxon
Nicotiana speciesHaploid chromosome number (n)Number of loci
N. sylvestris1231
N. obtusifolia1231
N. stocktonii2421
N. nesophilla2421
N. repanda2431
N. nudicaulis2421

Southern hybridization was used to reveal the entire population of 5S rDNA units. There is a near-identical hybridization pattern in all species of section Repandae with a predominant 5S rDNA band at 0.50 kb (Fig. 4). All species of section Repandae except for N. nudicaulis also have a faint minor band at 0.60 kb. This band is of similar size to the predominant band in section Trigonophyllae (N. obtusifolia, including N. palmeri). It is therefore possible that a few, unaltered Trigonophyllae units remain in section Repandae. The predominant N. sylvestris band is at 0.43 kb. The major family of 5S rDNA in section Repandae is dissimilar in size to both species likely to be the closest extant descendents of the two parents. However, the sequence of the predominant NTS type does indicate similarity to section Trigonophyllae.

It is likely that the 5S units in section Repandae are of Trigonophyllae origin, but there has been divergence in Repandae and Trigonophyllae since allopolyploid formation. This divergence must have been associated with inter- and intralocus homogenization because the vast majority of units in each species is identical in size. The uniformity of the 5S rDNA units across the four species of section Repandae is a surprise given that divergence of the diploid species of section Tomentosae is associated with the evolution, amplification and deletion of an array of 5S rDNA families (Matyasek et al., 2002). We suggest that the common ancestor of Repandae already possessed a repeat of this size before speciation took place or that it was inherited from a parental lineage that has subsequently diverged. It is unlikely that all four species could have independently developed repeats of the same size.

The patterns of 5S rDNA evolution in the allopolyploid section Repandae are different from those of the similarly allopolyploid N. tabacum (c. 0.2 Myr) in which 5S units of both N. sylvestris and N. tomentosiformis origins remain intact and occur at their expected chromosomal loci (Matyasek et al., 2002). Perhaps the different ages of the allopolyploids are an important factor. The differences between N. tabacum and the species of section Repandae are reflected elsewhere. For example, no 5S rDNA loci loss or interlocus sequence homogenization is observed in the 5S arrays of Gossypium allopolyploids (Cronn et al., 1996). These allopolyploids are approx. 1.5 Myr old and formed from parental diploids that are separated by 6.7 Myr of evolution (Senchina et al., 2003). However, in Glycine, Krishnan et al. (2001) found 5S rDNA locus loss in putatively recently formed polyploids. They showed that, as in Repandae, Glycine allotetraploids [e.g. G. tabacina (Labill.) Benth.] have been reduced to the same locus number as their ‘diploid’ progenitors.

It has been proposed that sequences located near chromosome ends are more prone to genetic change (e.g. gene conversion, elimination, translocation) than sequences at other positions (Kovarik et al., 2004). However, here in Repandae this does not explain the observed locus loss because it is the subtelomeric 5S rDNA of N. obtusifolia origin that has been retained, whereas the more proximal locus of N. sylvestris origin has been lost.

35S rDNA evolution in section Repandae

Three 35S rDNA loci can be detected in each of the two extant species most closely related to the progenitor diploid species of section Repandae. Therefore, we might expect six loci in allopolyploid Repandae, but species have either two or three 35S rDNA loci (Table 2). Chase et al. (2003) previously published a phylogenetic tree based on nuclear ITS sequences of Nicotiana. Here the tree is reproduced for the species studied in this work (Fig. 5). As noted by Chase et al. (2003), ITS sequences of Repandae are similar to those of N. sylvestris, which is the species most closely related to their maternal diploid progenitor. Southern hybridization reveals the gross unit structure for all populations of 35S rDNA. All species examined had a unique profile of bands, with the exception of N. stocktonii and N. nesophila, which are closely related species (i.e. they have only a few differences in their DNA sequences). However, the sequence similarity of repeats in each Repandae species indicates that there has been sequence divergence in sections Repandae and/or Sylvestres, sequence homogenization and locus loss. Previously we have shown in the allopolyploids N. tabacum, N. rustica and N. arentsii that the parental origin of the 35S rDNA loci does not seem to influence the direction of sequence homogenization (Kovarik et al., 2004). In section Repandae, inheritance/maintenance of 5S unit types that are similar to the paternal Trigonophyllae parent and of 35S units typical of the maternal Sylvestres parent (Fig. 5) demonstrate that inheritance patterns of these two different rDNA loci are not linked.

Figure 5.

The single most-parsimonious nontranscribed spacer (NTS) tree (left) vs the strict consensus of all most-parsimonious internal transcribed spacer (ITS) trees (right). Both were analyzed using identical parameters (see the Materials and Methods section). The striped clade indicates section Repandae.

The more recently formed allopolyploids –N. tabacum, N. arentsii and N. rustica– have the sum and location of the 35S units that would be expected by direct combination of their progenitor genomes. However, these units have been overwritten, at least to some extent, by one parental type (concerted evolution), summarized previously (Kovarik et al., 2004). In Nicotiana, this situation can be viewed as the ‘first level’ of diploidization of 35S rDNA loci. In section Repandae, the number of rDNA loci is reduced, and this can be considered as a ‘second level’ of rDNA diploidization.


We would like to thank Linda Jenkins and John Sitch for the cultivation of plants used in this study. Thanks to Martyn Powell, Dion Devey and David Springate for technical support in the laboratory. This research was funded by the Natural Environmental Research Council and Grant Agency of the Czech Republic (award 521/04/0775).