Chromosomal stasis in diploids contrasts with genome restructuring in auto- and allopolyploid taxa of Hepatica (Ranunculaceae)


Author for correspondence: Hanna Weiss-Schneeweiss Tel: +43 14277 54150 Fax: +43 14277 9541 Email:


  • • Polyploidization and chromosomal rearrangements are recognized as major forces in plant evolution. Their role is investigated in the disjunctly distributed northern hemisphere Hepatica (Ranunculaceae).
  • • Chromosome numbers, karyotype morphology, banding patterns, 5S and 35S rDNA localization in all known species were investigated and interpreted in a phylogenetic context established from nuclear internal transcribed spacer (ITS) and plastid matK sequences.
  • • All species had a chromosome base number of x = 7. The karyotype was symmetric and showed little variation among diploids with one locus each of 5S and 35S rDNA, except for interpopulational variation concerning 35S rDNA loci number and localization in H. asiatica. Tetraploids exhibited chromosomal changes, including asymmetry and/or loss of rDNA loci. Nuclear and plastid sequences resulted in incongruent topologies because of the positions of some tetraploid taxa. The diversification of Hepatica occurred not earlier than the Pliocene.
  • • Genome restructuring, especially involving 35S rDNA, within a few million yr or less characterizes evolution of both auto- and allopolyploids and of the diploid species H. asiatica, which is the presumptive ancestor of two other diploid species.


Chromosomal changes are recognized as a major factor accompanying plant speciation (Stebbins, 1971; Levin, 2002; Raskina et al., 2004; Lysak et al., 2006). These changes range from multiplications of whole chromosome sets with or without hybridization (allo- and autopolyploidy) to changes in numbers of single chromosomes (aneuploidy), and finally to structural changes involving one or more chromosomes (e.g. translocations, inversions, fusions). While chromosome numbers as well as their changes and basic information about karyotypes (chromosome length, position of centromeres, etc.) have long been used as additional sources for evolutionary inference, only the recent rapid development of powerful cytogenetic techniques allows tracing of the often subtle structural changes of chromosomes and interpretation of this information in a phylogenetic context (Clarkson et al., 2005; Lysak et al., 2006). Fluorescence in situ hybridization (FISH) allows characterization and mapping of individual families of both repetitive and single copy DNAs, such as rDNA, species/genus-specific repetitive DNA, and DNA from individual chromosomes or chromosome fragments (Kulikova et al., 2001; Cuadrado & Jouve, 2002; Mishima et al., 2002; Pedrosa et al., 2002; Lysak et al., 2006), while genomic in situ hybridization (GISH) allows hybridization of whole parental genomes to chromosomes of hybrids and allopolyploids (Schwarzacher et al., 1989; Takahashi et al., 1997; Hayasaki et al., 2000). Molecular cytogenetic methods are well established for plant chromosomes (Schwarzacher & Heslop-Harrison, 2000) and have already been successfully applied to address evolutionary questions mostly in groups of economic importance, such as Nicotiana (Solanaceae, Lim et al., 2000; Clarkson et al., 2005) and Beta (Chenopodiaceae, Schmidt & Heslop-Harrison, 1996; Gao et al., 2000; Dechyeva et al., 2003).

Hepatica (Ranunculaceae) is a small genus of about a dozen species disjunctly distributed in temperate zones of the northern hemisphere (Meusel et al., 1965; Tamura, 1995). Most species are diploids based on x = 7 (2n = 14: H. acutiloba and H. americana in the New World; H. nobilis var. nobilis in Europe; H. falconeri in central Asia; H. asiatica, H. insularis, H. maxima and H. nobilis var. japonica in eastern Asia), but a few taxa appear to be exclusively tetraploids (2n = 4x = 28; H. transsilvanica in Europe; H. henryi and H. yamatutai in China; H. nobilis var. pubescens in Japan). The origin of the polyploids is unclear, and both auto- and allopolyploidy have been invoked as explanations. In particular, based on morphological data, H. falconeri has been suggested as a putative parental species of H. transsilvanica, H. henryi and H. yamatutai, with the second parental species probably being an entire-leaved taxon (H. nobilis var. nobilis in the case of H. transsilvanica, and east Asian taxa, in particular the most widespread H. asiatica, in the case of H. henryi and H. yamatutai). By contrast, H. nobilis var. pubescens has been hypothesized to be of autopolyploid origin from H. nobilis var. japonica (Mabuchi, 1998). As chromosome numbers and basic karyotypes show little differentiation among taxa, more powerful techniques are necessary to test hypotheses regarding the origin of the polyploids.

While polyploidization apparently played an important role in diversification within Hepatica, differentiation at the diploid level is poorly understood, especially in eastern Asia, where several diploid taxa occur. Specifically, H. maxima, endemic to Ullung Island off the coast of Korea, is suggested to be a derivative of the Korean mainland species H. asiatica (Pfosser et al., in press). Additional complications arise from the lack of reciprocal monophyly between H. asiatica and the South Korean endemic H. insularis (Pfosser et al., in press). A sequence type that is known to participate actively in genome rearrangements is rDNA (Schubert & Wobus, 1985; Clarkson et al., 2005), and thus analysing and comparing rearrangements of 5S and 35S rDNA potentially allow testing of hypotheses on relationships between these three species.

The aims of this study were to summarize previously published and newly obtained chromosome numbers and karyotype information in species of Hepatica for analysing the overall variation in karyotype structure and karyotype changes that accompany polyploidization. These data are combined with information obtained from FISH with 5S and 35S rDNA to test the hypothesized origins of the polyploid taxa and to infer the relationships among the east Asian endemics H. asiatica, H. insularis and H. maxima. In a complementary approach, sequence data from the nuclear and plastid genomes are used to infer phylogenetic relationships within the genus, with special emphasis on the polyploid taxa, and to obtain age estimates for infrageneric diversification.

Materials and Methods

Plant material

Material of all Hepatica taxa except H. americana was collected in the field and, except for H. falconeri and H. yamatutai, grown in the Botanical Garden of Vienna. Information on voucher specimens is given in Table 1. For the karyological and cytological investigations, actively growing root-tip meristems were pretreated with 0.1% colchicine for 24 h at 4°C, fixed in ethanol : acetic acid (3 : 1) for 12 h at room temperature, and stored at −20°C until use. Plant material for DNA sequencing was dried and stored in silica gel.

Table 1.  Plant material, voucher information and new chromosome numbers (2n) for Hepatica
TaxonLocality, collectora, voucher number (herbarium)2nb
  • a

    BYS, B.-Y. Sun; HK, H. Kato; KJK, K.-J. Kim; MP, M. Pfosser; TKP, T.-K. Paek; TS, T.F. Stuessy.

  • b

    Superscript numbers refer to the numbers of individuals analysed.

acutiloba DC.USA, Illinois, Lake Bluff, across from 220 Moffett Road, on slope of ravine, 17 August 2000: TS H25 (WU)141
asiatica NakaiKorea, Kyeungbuk Province, Yeung-Yang Co., Il-weol Myeon, Mt. Il-weol, 1 May 2001: TS, MP & TKP 17631 (WU, JNU)142
—, Chonbuk Province, Naejangsan National Park, c. 10 km S of Chongju City, 25 April 2001: TS, BYS, MP & TKP 17505 (WU, JNU)143
—, Kyeungbuk Province, Yeung-Yang Co., Il-weol Myeon, Mt Il-weol, 1 May 2001: TS, MP & TKP 17626 (WU, JNU)143
—, Kwangwon Province, forest near Sangwon-temple, Mt Odae, BYS & KJK, 20 December 1998 3220 (JNU)141
falconeri (Thomson) StewardPakistan, Nathiagali, Hazara, 16 May 2001: TM TM2 (herb. Mabuchi)141
Pakistan, Shogran, Hazara, 16 May 2001: TM TM5 (herb. Mabuchi)141
henryi StewardPurchased from garden company, TS H26 (WU)282
insularis NakaiKorea, Cheju Province, Cheju Island, Hallasan National Park, 26 April 2001: TS, BYS, MP & TKP 17516 (WU, JNU)142
—, —, —, 2 km S of Songpanak. 26 April 2001: TS, BYS, MP & TKP 17526 (WU, JNU)143
maxima (Nakai) Nakai—, Gun Province, Ullung Island, Taeha-ri, Seodaryong, 29 April 2001: TS, BYS, MP & TKP 17552 (WU, JNU)142
—, —, —, Namyang-ri, Tonqgumi nem, valley along the road, 30 April 2004: TS, BYS, MP & TKP 17606 (WU, JNU)142
nobilis Schreb. var. japonica NakaiJapan, Nagano Prefecture, Kawaji, Iida city, 1 June 2000: TS, BYS & HK 17329 (WU, JNU)141
—, —, —, —, 50 m further from 17329, 1 June 2000: TS, BYS & HK 17330 (WU, JNU)143
nobilis Schreb. var. nobilisAustria, Vienna, Wienerwald, March 2001: TS & BYS H24 (WU)142
Romania, garden material from Botanical Garden in Bucharest, TS H29 (WU)141
nobilis Schreb. var. pubescens (M. Hiroe) Kitam.Japan, Shiga Prefecture, Mt. Ibuki, up road of limestone quarry company, 900 m 2 June 2000: TS, BYS & HK 17333 (WU, JNU)283
—, —, —, —, 800 m 2 June 2000: TS, BYS & HK 17334 (WU, JNU)282
transsilvanica FussAustria, garden material from HBV, University of Vienna. TS H21 (WU)282
Romania, garden material from Botanical Garden in Bucharest. TS H30 (WU)281
yamatutai NakaiChina, Sichuan Province, Mt. Emei, purchased from garden company, TM TM1001 (herb. Mabuchi)281

Chromosome analyses

Feulgen staining with Schiff's reagent was performed according to standard protocols (Weiss et al., 2002). Squash preparations were made in a drop of 45% acetic acid. From each accession, a preparation with at least 10 well-spread chromosome plates was chosen for analysis and measurement. Idiograms were prepared using Autoidiogram software (courtesy of Dr Wolfgang Harand, University of Vienna).

For fluorochrome banding and FISH, chromosomes were prepared by enzymatic digestion/squashing as described by Schwarzacher & Heslop-Harrison (2000) and Weiss & Maluszynska (2000), with minor modifications. Briefly, material was digested with 1% (w/v) cellulase Onozuka (Serva, Heidelberg, Germany), 0.4% (w/v) cytohelicase (Sigma-Aldrich, Vienna, Austria), and 0.4% (w/v) pectolyase (Sigma-Aldrich) for 45–60 min at 37°C. Squash preparations were made in a drop of 60% acetic acid. The quality of spreads was checked in phase-contrast and only preparations with adequate numbers of well-spread metaphases (10–15) were selected for FISH. Slides were frozen at −80°C, and after cover-slip removal, preparations were stored at −20°C.

Chromomycin (CMA3) fluorescent banding was done following Schweizer (1976) and Schweizer & Ambros (1994), with slight modifications. The slides were incubated in McIlvaine buffer, pH 7, for 15 min, stained with 0.1 g l−1 CMA3 (Sigma-Aldrich) for 7 min, rinsed in the buffer and stained with DAPI (4′,6-diamidino-2-phenylindole; Sigma-Aldrich) for 7 min. After the final rinse, the slides were mounted in glycerol : McIlvaine (1 : 1) antifade buffer and incubated at 37°C for 2–3 d. Preparations were examined in an Axioplan2 epifluorescent microscope, and the images were acquired with a CCD camera and Axiovision 3.5 software (Carl Zeiss, Jena, Germany).

C-banding was performed according to standard protocols (Fukui & Nakayama, 1996) with minor modifications concerning the final staining of chromosomes, which was carried out with DAPI instead of Giemsa solution. Examination of preparations and image acquisition were performed as described above.

Fluorescence in situ hybridization was carried out following the methods of Schwarzacher & Heslop-Harrison (2000) and Weiss-Schneeweiss et al. (2003) with minor modifications. The chromosome preparations were pretreated with RNaseA (100 µg ml−1; Sigma-Aldrich) and pepsin (5 µg ml−1; Sigma-Aldrich) before hybridization. The hybridization mixture contained 50% formamide, 10% dextran sulphate, salmon sperm blocking DNA (0.1 µg µl−1; Sigma-Aldrich), 2 × SSC and 2 ng µl−1 of labelled probe. Probes used for FISH were: 18S and/or 25S rDNA from Arabidopsis thaliana, both in plasmid pSK+, and 5S rDNA from Beta vulgaris in plasmid pBx1-2 directly labelled with Cy3 (GE Healthcare, Chalfont St Giles, UK). 18S and 25S rDNA probes were labelled with digoxygenin using a nick translation kit (Roche Diagnostics, Mannheim, Germany), whereas 5S rDNA was labelled by PCR using standard M13 primers and Cy3-labelled nucleotides. Chromosome preparations, after applying the hybridization mix, were denatured on a hot plate (Hybaid Thermal Cycler PCR-in situ, Hybaid Ltd, Middlesex, UK) at 78°C for 4 min. Hybridization was allowed for 2 d. Stringent washes were performed at 40°C in the following solutions: 2 × SSC, 0.1 × SSC and 2 × SSC (5 min each). Digoxygenin-labelled 18S/25S rDNA probes were detected with antidigoxygenin antibody conjugated with FITC (Roche Diagnostics). The preparations were mounted in antifade buffer Vectashield (Vector Laboratories, Peterborough, UK) containing DAPI counterstain (2 µg ml−1) and stored at 4°C. Analyses of preparations and image acquisition were performed as described above. Contrast and colour balance of the images were adjusted using Corel PhotoPaint 10.0 (Corel, Ottawa, Canada) with only those functions that applied equally to all pixels in the image. For rDNA localization, an average of 15 well-spread metaphases and prometaphases was analysed for each species.

Molecular phylogenetic analyses

Total genomic DNA was isolated from silica gel-dried or air-dried leaf material as described in Pfosser et al. (in press). PCR amplification of partial matK and part of the flanking trnK intron was conducted using primers trnK345f, matK1360r, trnK3ar and trnK3r, and PCR conditions as described in Paun et al. (2005). PCR amplification of ITS sequences of several species not available from GenBank was performed according to the method described in Paun et al. (2005). Sequencing was performed using BigDye terminator technology (Applied Biosystems, Foster City, CA, USA) following the manufacturer's instructions, using the PCR primers and additionally, in the case of matK sequences of H. henryi and H. falconeri, an internal primer (5′-ACGATCAACGCTAGAG-3′). Sequences were visualized on an ABI 377 automated sequencer (Applied Biosystems).

Newly obtained sequences together with sequences downloaded from GenBank (Table 2) were aligned by eye using BioEdit 5.0.9 (Hall, 1999), designating Pulsatilla cernua (ITS, AB120213; matK, AB110531) and Anemone flaccida (ITS, AY055391; matK, AB110530) as outgroups. Initial analysis of the plastid data set suggested that H. americana (GenBank, AF542590; Hilu et al., 2003) would be unexpectedly distinct from the other Hepatica species. BLAST searches indicate that sequence positions 1–795 are nearly identical to sequences of other Hepatica species, while the sequence after position 796, which is given as ambiguous (N) in the GenBank entry, is up to 99% identical with matK sequences from Magnoliaceae, especially Liriodendron (E-value = 0.0). This suggests that the matK sequence of H. americana is actually chimerical, probably artificially in the course of assembling, and therefore we only included the first Hepatica-like part in further analyses.

Table 2.  List of taxa, voucher information (or reference for already published sequences) and GenBank accession numbers
TaxonGenBank accessions (voucher numbera/reference)
  1. a See Table 1 for details.

Anemone flaccida Fr. SchmidtAY055391 (Schuettpelz et al., 2002)AB110530 (Miikeda et al., 2006)
Pulsatilla cernua (Thunb.) Bercht. & J. PreslAB120213 (Miikeda et al., 2006)AB110531 (Miikeda et al., 2006)
H. acutilobaAM267285 (Pfosser et al., in press)DQ994677 (H25)
H. americanaAY055386 (Schuettpelz et al., 2002)AF542590 (Hilu et al., 2003)
H. asiaticaAM267289 (Pfosser et al. in press)DQ994672 (17626)
H. falconeriEF012263 (TM2)DQ994675 (TM2)
H. henryiAM267290 (Pfosser et al., in press)DQ994674 (H26)
H. insularisAM267288 (Pfosser et al., in press)DQ994673 (17506)
H. maximaAM267282 (Pfosser et al., in press)DQ994671 (17552)
H. nobilis var. japonicaAM267286 (Pfosser et al., in press)AB110532 (O. Miikeda et al., 2006)
H. nobilis var. nobilisAM267284 (Pfosser et al., in press)DQ994668 (H24), DQ994669 (H29)
H. nobilis var. pubescensAM267287 (Pfosser et al., in press)DQ994667 (17334)
H. transsilvanicaAM267283 (Pfosser et al., in press)DQ994670 (H21)
H. yamatutai NakaiEF012264 (TM1001)DQ994676 (TM1001)

Maximum-parsimony analysis, employing the branch-and-bound search with the furthest sequence addition and treating characters as unordered, and maximum-likelihood analysis, using starting trees obtained via neighbour-joining and TBR branch swapping with iteratively optimized model parameters (see Supplementary Material for details), were conducted using paup* 4.0b10 (Swofford, 2001). The best-fit substitution models (determined using the Akaike Information Criterion as implemented in Modeltest 3.7; Posada & Crandall, 1998) are a general time-reversible model with a gamma distribution accounting for heterogeneity among sites (GTR + Γ) for the ITS data set and a two-transversion-parameter model with unequal base frequencies and a proportion of invariable sites (K81uf + I) for the plastid data set. Clade support was assessed using bootstrap with 1000 replicates with the same search options. Bayesian analysis was performed with MrBayes 3.1 (Ronquist & Huelsenbeck, 2003) using models with six substitution types (nst = 6) and a gamma distribution for describing rate heterogeneity across sites (six and 12 categories for the ITS and plastid data set, respectively), estimating values for all parameters during analysis, and employing three runs with four chains each for 2 × 106 generations, each with trees being sampled every 250th generation using the default priors (see Supplementary Material). Convergence of the independent runs was assessed as described in Park et al. (2006) and the posterior probability (PP) of the phylogeny and its branches was determined from the posterior set of 21 600 trees.

Alternative hypotheses were tested in a maximum-likelihood framework using the Shimodaira–Hasegawa test (SH-test, Shimodaira & Hasegawa, 1999) as implemented in paup* 4.0b10, employing 10 000 bootstrap replicates to generate a test distribution by the RELL (re-sampling estimated log-likelihood) method. The strategy for the constrained maximum-likelihood searches was the same as for the unconstrained searches described earlier. Because the SH-test is known to be very conservative (i.e. less likely to reject the null hypothesis of all trees in the set being equally good explanations of the data; Goldman et al., 2000), we also used a Bayesian approach by determining the posterior probabilities of alternative topologies from the combined set of trees after the burn-in period (discussed earlier), considering alternative topologies with posterior probabilities of 0.05 or more as not significantly worse (Huelsenbeck et al., 2002; Steele et al., 2005).

There is no direct fossil evidence for the age of the genus Hepatica. Therefore, as a first approach, we use published substitution rates of 8.34 × 10−9 − 3.89 × 10−8 substitutions per site and year for ITS (Zhang et al., 2001; see Park et al., 2006 for a more detailed discussion of substitution rates for ITS) and 2.2 × 10−9 ± 0.68 × 10−9 substitutions per site and year for matK (Paun et al., 2005) to obtain age estimates for the crown-group age of Hepatica. We tested for clock-like evolution of substitution rates in the two data sets using likelihood ratio tests (Huelsenbeck & Rannala, 1997). If the null hypothesis of a molecular clock is not rejected, branch lengths can be directly translated into divergence times. Otherwise, we used the longest edge within Hepatica, blmax, to derive a conservative estimate of the divergence time, t, as described in Park et al. (2006). Owing to the sparse sampling among outgroups, we do not provide estimates for the stem-group age. As a second approach, we used relaxed molecular clock methods implemented in the program r8s (version 1.7; program available from the author:, namely nonparametric rate smoothing (NPRS; Sanderson, 1997) and penalized likelihood (PL; Sanderson, 2002), as described in Park et al. (2006). The optimal values for smoothing parameter for PL analysis were determined to be 0.16 and 0.1 for the ITS and matK data set, respectively. As external calibration point we used the age of Ullung Island (off the coast of South Korea), determined to have emerged above water c. 1.8 million yr ago (Mya) (Kim, 1985a,b) for constraining the age of the island endemic H. maxima. For computational reasons we fixed the stem node of H. maxima at this age. The implementation of ages derived from the emergence of oceanic islands is not simple, as the species now endemic to the island might have originated before the island's emergence (in which case the age of the island provides a minimum age) or long after it (in which case the age of the island provides a maximum age; Heads, 2005; Renner, 2005). Therefore, all the age estimates obtained must be interpreted with the necessary caution.


Chromosome numbers

Chromosome counts newly obtained in this study (Table 1) agree with the previously reported distribution of chromosome numbers and ploidy levels in Hepatica (see Table S1, Supplementary Material). All taxa have the same basic chromosome number of x = 7: H. acutiloba, H. asiatica, H. falconeri, H. insularis, H. maxima, H. nobilis var. nobilis and var. japonica are diploid (2n = 2x = 14) and H. henryi, H. nobilis var. pubescens, H. transsilvanica and H. yamatutai are tetraploid (2n = 4x = 28). The only exceptions are reports of 2n = 16 for H. acutiloba (Sugiura, 1931, 1936; published as A. hepatica var. acuta) and H. transsilvanica (Pop, 1937; Tǎcinǎ, 1980). However, x = 8 has never been reported again for any Hepatica species, and we consider these numbers to be erroneous, probably as a result of confusion with material from an Anemone species with x = 8. Each taxon is uniform concerning its ploidy level (Table S1; H. Weiss-Schneeweiss, unpubl.), with a few exceptions. For the two Japanese taxa, H. nobilis var. japonica (2x) and var. pubescens (4x), higher ploidy levels of 4x and 6x, respectively, have been reported (Hiroe, 1957; Kondo, 1964). Tetraploid H. nobilis var. japonica is difficult to verify and might be the result of misidentification, while the hexaploid H. nobilis var. pubescens has never been confirmed, although plants for newer analyses have been collected in the same locality (Hara & Kurosawa, 1958; Mabuchi, 1980, 1998). The third exception concerns the reports of triploid H. nobilis var. japonica (Suda, 1975; Mabuchi, 1998) and H. transsilvanica (2n = 3x = 21; Baumberger, 1970), the latter count being obtained from material provided by a garden supply company, which might have included vegetatively propagated hybrid individuals with diploid H. nobilis. Mabuchi (1998), however, found a natural triploid individual which probably originated from a cross between H. nobilis var. japonica and H. nobilis var. pubescens.

General karyotype structure and banding patterns

Karyotypes of diploid Hepatica species are uniform with six pairs of meta- to submetacentric chromosomes and one pair of subtelocentric chromosomes, the latter carrying satellites on their short arms. As chromosome measurements depend greatly on chromosome condensation, the assignment of chromosomes to pairs one to five is difficult and is in part arbitrary because of the lack of pronounced differences in their lengths and type. In some species, chromosome pair number 6 is more clearly submetacentric than the others, and therefore it can be usually distinguished more easily from other chromosome pairs (Figs 1, 2). This is of particular importance because this chromosome pair carries the 5S rDNA locus in all diploid species. In H. transsilvanica, two types of chromosome 6 can be distinguished, differing in the length of the short arm (also discussed later).

Figure 1.

Mitotic metaphases (a–l, p–t) and interphase nuclei (m–o) of Hepatica species (for details on accession numbers, see Table 1) showing fluorescence in situ hybridization signals (red and green signals correspond to 5S and 35S rDNA loci, respectively) (a–o), C-banding patterns (p), and fluorescent banding with chromomycin (CMA3) (q, s) and 4′,6-diamidino-2-phenylindole (DAPI) (r, t). (a) H. acutiloba H25; (b) H. asiatica 17626; (c, p) H. asiatica 3220; (d) H. falconeri TM2; (e) H. insularis 17526; (f) H. maxima 17552; (g) H. nobilis var. japonica 17330; (h, q, r) H. nobilis var. nobilis H24; (i, o) H. henryi H26; (j, n) H. nobilis var. pubescens 17334; (k, s, t) H. transsilvanica H30; (l) H. transsilvanica H21; (m) H. yamatutai TM1001. Bar, 10 µm.

Figure 2.

Idiograms of Hepatica species. Positions of 5S and 35S rDNA genes are marked as open circles and black circles/blocks, respectively.

Application of the C-banding technique allows localization of heterochromatin blocks. In diploid species of Hepatica, heterochromatic blocks are localized only in subtelomeric positions, and in some species additionally as minute dots in regions corresponding to rDNA loci (chromosomes 6 and 7; Fig. 1). No differences in localization of subtelomeric blocks are observed among the populations analysed. In polyploids, localization of heterochromatin is similar, and subtelomeric blocks are present in all species analysed (data not shown). In H. nobilis var. pubescens, additional dot-like heterochromatic signals are observed in chromosome 5 and correspond to an additional locus of 35S rDNA (discussed later).

Fluorescence banding (CMA3 and DAPI) allows distinguishing of AT- from GC-rich heterochromatic regions. CMA3/DAPI banding of Hepatica chromosomes reveals the presence of AT-rich heterochromatin in the subtelomeric regions of all chromosomes except for the subtelomeric region of the short arm of chromosome 7 (Fig. 1r), thus corresponding to the majority of heterochromatin blocks detected by C-banding (Fig. 1p). GC-rich blocks correspond to localization of the 35S rRNA genes (Fig. 1q,s). No variation is observed among diploids, and the slight variation in polyploids corresponds to variation in number and distribution of rDNA loci (discussed later).

5S and 35S rDNA localization

In all diploid taxa (H. acutiloba, H. asiatica, H. falconeri, H. insularis, H. maxima, H. nobilis var. japonica and H. nobilis var. nobilis), at least one locus each of 35S rDNA and 5S rDNA is present. The latter is localized close to the centromere within the short arm of chromosome 6 (Fig. 1a–l). Although this chromosome shows slight variation among species in the length of the short arm, the 5S rDNA locus is located at a similar distance from the centromeric region and is of equal signal size/strength in all species. 35S rDNA is localized within the short arm of chromosome 7, forming a satellite (Fig. 1a–l). This region is also positive in C-banding and is GC-rich (CMA3-positive, DAPI-negative; Fig. 1q–t). The size of the satellite, and thus the number of repeats in each 35S rDNA locus, varies both among species and among individual plants, and even between homologous chromosomes in one cell (Fig. 1h–l). In H. asiatica, interpopulational variation in 35S rDNA loci number and localization is observed. Two out of four analysed populations (accessions 17505 and 17626; Table 1) have only the locus on chromosome 7 (Fig. 1b), while in the two other populations (accessions 3220 and 17631; Table 1) an additional 35S rDNA locus is present within the short arm of chromosome 6 in close proximity to the 5S rDNA locus, but more distally within the chromosome arm (Fig. 1c).

More variation in number and/or localization of rDNA loci is observed in the four tetraploid taxa analysed. In H. henryi, two out of four chromosomes 6 reveal a lower short : long arm ratio than the other two, and only those have 5S rDNA locus located in their short arms (Figs 1i and 2). All four telocentric chromosomes 7 carry satellite and a 35S rDNA locus on the short arm, but two of them have a stronger FISH signal (Figs 1i, 2). The scarcity of material of H. yamatutai did not allow good-quality images of chromosomes after FISH, but the number of signals of both 5S and 35S rDNA in the interphase nuclei (Fig. 1m) agrees with the number and size classes of signals seen in chromosomes of H. transsilvanica (Fig. 1k), especially when 5S rDNA is considered. In H. nobilis var. pubescens, 5S rDNA is localized proximally on the short arms of all four chromosomes 6 (Figs 1j, 2). As in H. henryi, the 35S rDNA locus in the short arms of four chromosomes 7 differs in signal strength, two signals being much stronger than the other two (Figs 1j,n, 2). An additional 35S rDNA locus is found in the long arms of two out of four chromosomes 5 and is unique for this species (Figs 1j, 2). The 5S rDNA locus in H. transsilvanica is localized on the short arms of all four chromosomes 6, but again two of the sites either have lower FISH signal intensity (Figs 1k, 2) or are absent/undetected (Fig. 1l). The four chromosomes 6 differ in their structure (Fig. 2). The signal intensity of 35S rDNA located within the satellite on the short arm of all four chromosomes 7 varies in individual chromosomes and individuals.


Different phylogenetic methods result in topologies differing only slightly in the degree of resolution and partly in nodal support (Fig. 3). Parameters of the different data sets, none of which includes polymorphic sites, and of different phylogenetic analyses are summarized in Table 3. Phylogenetic relationships inferred from the plastid marker matK are less well resolved than from nuclear ITS. Apart from some groups, which are congruently inferred from, or at least not contradicted by, any of the two markers, for instance the clade formed by the two New World taxa H. americana and H. acutiloba or the group comprising H. asiatica, H. henryi, and H. insularis, others are contradictory (Fig. 3). These contradictions result in incongruences of the data sets, if analyses of the ITS data are constrained to the topology inferred from the plastid data (SH-test: –lnconstrained = 1570.86202, δ ln = 15.94254, P = 0.0368), but not in the other direction (SH-test: –lnconstrained = 2393.92084, δ ln = 18.91954, P = 0.0551). To assess if these incongruences could be caused by the putatively allopolyploid taxa H. henryi, H. yamatutai and H. transsilvanica, the SH-test was repeated based on trees obtained from the reduced data sets. The results of these tests no longer suggest incongruence between the nuclear and the plastid data (SH-test: –lnunconstrained = 1455.79467, –lnconstrained = 1466.50599, δ ln = 10.71131, P = 0.0614) and strengthen the rejection of incongruence between the plastid and the nuclear data set (SH-test: –lnunconstrained = 2307.41717, –lnconstrained = 2313.54774, δ ln = 6.13057, P = 0.1245). Excluding additionally H. nobilis var. pubescens gives nearly identical results (data not shown), as is expected from the congruent phylogenetic position in both data sets and the low amount of sequence divergence between H. nobilis var. pubescens and H. nobilis var. japonica. Using the Bayesian approach for hypothesis testing, the two data sets are inferred as incongruent in both directions with and without polyploids (P < 0.001). Although not inferred by any method used, a monophyletic H. nobilis is rejected by both markers only by the Bayesian hypothesis-testing approach (P = 0.002 and P < 0.001 for ITS and matK, respectively), but not by the maximum-likelihood-based SH-test (–lnconstrained = 1559.32551, δ ln = 4.40603, P = 0.1693; –lnconstrained = 2347.31446, δ ln = 5.34345, P = 0.1572).

Figure 3.

Phylogenetic relationships of Hepatica inferred from maximum likelihood of the nuclear internal transcribed spacer (ITS; left) and the plastid matK region (right). Numbers at the branches are (from top) bootstrap values from maximum parsimony, bootstrap values from maximum likelihood, and posterior probabilities ≥ 0.5 from Bayesian inference (in the case of nodes not supported by all methods, the respective missing support values are indicated by ‘n.a.’). Bar indicates substitutions/site (note that the branches for the outgroup taxa in the ITS phylogram are truncated for graphical reasons).

Table 3.  Parameters of different data sets and phylogenetic analyses
  • a

    Hepatica only.

  • b

    Harmonic mean.

Sequence lengtha (base pairs) 576–5771380
Number of characters
 All 5871384
 Variable 127  75
 Parsimony-informative  32  19
Number of MP trees  18   4
Tree length 150  81
Consistency index   0.9200   0.9753
Retention index   0.7551   0.9200
Rescaled consistency index   0.6947   0.8973
–ln score maximum likelihood1554.919492341.97101
–ln score Bayesian inferenceb1595.832389.44

Age estimation

A model of clock-like evolution is clearly rejected for the ITS sequences (–lnno-clock = 1554.91949, –lnclock = 1576.77252, 2 × δ ln = 43.70606, P < 0.001, 12 d.f.), but not for the matK sequences (–lnno-clock = 2341.97101, –lnclock = 2352.07845, 2 × δ ln = 20.21488, P = 0.0899, 13 d.f.). Using the longest edge in the ITS maximum-likelihood topology (the one leading to H. transsilvanica: Fig. 3), the crown group age of Hepatica is estimated to lie between (4.3–)2.5(−0.8) and (0.9–)0.5(−0.2) Mya. Similar age estimates are obtained from the plastid data, if the longest edge in the matK maximum-likelihood topology (that leading to H. nobilis var. pubescens) is chosen, covering (4.1)2.8–0.4(0.3) Mya, compared with an age of (1.4–)1.0(−0.8) Mya when using the clock-constrained topology. The respective age estimates obtained from NPRS/PL analyses are 2.84/3.35 and 2.57/3.06 Mya for the nuclear and plastid data, respectively.


The diploid members of the genus Hepatica (H. acutiloba, H. americana, H. asiatica, H. insularis, H. maxima, H. nobilis var. nobilis and var. japonica) are very uniform in chromosome number (x = 7), karyotype structure, heterochromatin distribution and rDNA localization (one locus each of 5S and 35S rDNA). This chromosomal stasis (except for variation seen in H. asiatica, see the following paragraph) agrees with the young age of the genus inferred from molecular clock analyses. Although the wide and disjunct distribution area of Hepatica in temperate regions of the northern hemisphere is suggestive of an ancient origin of this group, it appears that the diversification took place not earlier than the Pliocene, and at least partially in the Pleistocene (age estimates ranging from 4 to < 0.5 Mya). The stem group age of Hepatica, that is, its divergence from the group of Anemone species also with x = 7 (subg. Anemonidium, in contrast to subg. Anemone with x = 8: Ehrendorfer & Samuel, 2001; Schuettpelz et al., 2002; Pfosser et al., in press), might be considerably older, but the sampling outside Hepatica does not allow this question to be addressed.

In Hepatica, the only diploid species variable with respect to rDNA localization is H. asiatica. While two of four accessions analysed have only one 35S rDNA locus on chromosome 7, as in all other diploid taxa, two accessions have an additional 35S rDNA locus on chromosome 6, adjacent to the 5S rDNA locus (Fig. 4). The most parsimonious explanation is that parts of the rDNA locus from the satellite region of chromosome 7 have been translocated to chromosome 6. Both populations are geographically close (Pfosser et al., in press), which suggests that this chromosomal rearrangement originated once. Hepatica asiatica is a morphologically variable species (T. F. Stuessy & B.-Y. Sun, unpubl.), which gave rise to H. maxima, endemic to Ullung Island, and to H. insularis, restricted to the southern Korean peninsula and several Korean islands (Pfosser et al., in press). The karyological and cytological data indicate that both species originated from the typical chromosomal race of H. asiatica, and the aberrant chromosomal race of H. asiatica reported here could reflect still another, but more recent line of divergence. It has been suggested for eukaryotes in general that chromosomal rearrangements, if not causing speciation, can at least potentially intensify existing reproductive barriers (Coghlan et al., 2005), and thus H. asiatica might be an example for incipient speciation. Further studies on the distribution of cytotypes and the behaviour of intercytotype crosses are needed to address this question.

Figure 4.

Hypothetical major pathways of 5S (open circles) and 35S rDNA loci (closed symbols) evolution in taxa of Hepatica. Only those chromosomes carrying rDNA loci (chromosomes 5–7) are shown.

In contrast to the diploids, the polyploid taxa (H. henryi, H. nobilis var. pubescens, H. transsilvanica, H. yamatutai) are more variable in both basic chromosome morphology and nuclear DNA content, as well as in number and intensity of rDNA signals. Changes in rDNA distribution, which are congruently detected in different individuals and populations from the same taxon, can be classified as follows (Fig. 4): (i) diploidization of 5S rDNA on chromosome 6 in putative allopolyploids via reduction (H. transsilvanica, H. yamatutai) and eventual loss (H. transsilvanica, H. henryi) of two out of four sites, suggesting perhaps bivalent pairing of chromosomes of parental taxa; (ii) diploidization of 35S rDNA locus on chromosome 7 via gradual loss of two out of four sites in all polyploid taxa; (iii) occurrence of an additional locus of 35S rDNA on two out of four chromosomes 5 in putative autopolyploid H. nobilis var. pubescens. Thus, a general trend appears to be diploidization of both 5S and 35S rDNA loci after polyploidization, although to a different extent and via different modes (deletion, conversion) in different taxa. In allopolyploid taxa of Nicotiana, the degree of rDNA diploidization roughly agrees with the age of the polyploids, being advanced in species of sect. Repandae (c. 5.4 Myr old; Clarkson et al., 2005) but not in N. tabacum (c. 0.2 Myr old; Lim et al., 2004). Similarly, different extents of diploidization in Hepatica might reflect different ages of the polyploids, but the currently available data do not allow the development of a sufficiently detailed temporal framework for the evolution of the polyploids. While intra- and interindividual size variations of the satellite region on homologous chromosomes have been observed in the diploid H. nobilis var. japonica (Suda, 1975), polyploid taxa always show the same pattern of satellite size asymmetry (and rDNA copy number distribution; Mabuchi et al., 1987; this study). This suggests that the changes in polyploids are directional and fixed. The extent of genomic rearrangements observed in Hepatica (rDNA diploidization and translocation) lies well within the range observed in other taxa, ranging from more or less additive patterns – albeit partly with changes as a result of concerted evolution and/or sequence elimination (Kovarik et al., 2005; Skalickáet al., 2005) – in recent polyploids (Gossypium spp., Liu et al., 2001; Spartina anglica, Ainouche et al., 2003; Tragopogon spp., Pires et al., 2004) to larger genomic restructurings in usually much older polyploids (wheat, Ozkan et al., 2001; Nicotiana sect. Repandae, Clarkson et al., 2005). 35S rDNA is known for its mobility in the genome and has been reported to participate actively in genome rearrangements in some plant groups (Schubert & Wobus, 1985; Hall & Parker, 1995; Weiss-Schneeweiss et al., 2003). Its uniformity in diploid species and, with the exception of H. nobilis var. pubescens, the lack of detectable changes in chromosomal positions in polyploids indicate that this mode of genome evolution is of relatively minor importance in the evolution of Hepatica.

Several hypotheses have been put forward for the origin of the polyploid Hepatica taxa, but changes in karyotype structure and in rDNA distribution are inconclusive in this respect. Based on morphological (entire leaves), molecular, meiotic chromosome pairing and genome size data (Mabuchi, 1998; Kokubun et al., 2004; Mabuchi et al., 2005; Pfosser et al., in press), H. nobilis var. pubescens is suggested as an autotetraploid derivative of the diploid H. nobilis var. japonica. This is supported by the high degree of similarity of the four homologues of all chromosome pairs, including pairs 6 and 7 that carry rDNA, but it is at odds with the occurrence of an additional 35S rDNA locus on chromosome 5, which otherwise could be interpreted as evidence for an allopolyploid origin. No diploid species of Hepatica, however, possesses the additional 35S rDNA locus on chromosome 5, and therefore an alternative hypothesis, that is the occurrence of an additional 35S rDNA locus as the result of genomic changes subsequent to polyploidization, is more likely. Although autopolyploidy is no longer considered an evolutionary dead end (Stebbins, 1971), but is appreciated as an important process in plant evolution (Ramsey & Schemske, 1998, 2002; Wendel, 2000), putative genomic changes are less well understood than in the comparatively better investigated allopolyploids, such as Brassica or wheat (Song et al., 1995; Wendel, 2000; Ozkan et al., 2001). Investigations in South American species of Hypochaeris (Asteraceae; Weiss-Schneeweiss et al., 2003; H. Weiss-Schneeweiss, unpubl.) indicate that autopolyploids undergo rapid (and partly directional) rearrangements involving rDNA, and a similar pattern is observed here in H. nobilis var. pubescens.

In contrast to H. nobilis var. pubescens, the three remaining polyploids, H. henryi, H. transsilvanica and H. yamatutai, have crenate leaves, a feature shared only with the central Asian diploid, H. falconeri. This prompted Mabuchi et al. (2005) to propose that the three polyploids are autopolyploid derivatives of H. falconeri (or a related extinct crenate-leaved taxon). Alternatively, the presence of crenate leaves in the tetraploids could be explained by their allopolyploid origin involving H. falconeri as one of the parental species. However, karyotypically, H. falconeri is identical to the other diploid taxa (Ogisu et al., 2002; this study), rendering any inferences from this character impossible. Genome size data (Mabuchi et al., 2005), however, seem rather to support the hypothesis of allopolyploid origin. The nuclear DNA amount determined for H. henryi, H. transsilvanica and H. yamatutai agrees well with an essentially additive pattern of parental genome size, whereas the alternative hypothesis of autopolyploid origin requires additional and independent increase of genome size after autopolyploidization. This second hypothesis is not only less parsimonious, but it would also imply a deviation from the general trend in angiosperms of genome size reduction after polyploidization (Kellogg & Bennetzen, 2004; Leitch & Bennett, 2004; Weiss-Schneeweiss et al., 2006). In both nuclear and plastid data (Pfosser et al., in press; this study), H. henryi is associated with H. asiatica and H. insularis, although their relationships are not always resolved (Fig. 3). Hepatica asiatica shows an unusual amount of intraspecific variability in rDNA localization (this study) and it has been suggested to be an ancestor of both H. insularis and H. maxima (Pfosser et al., in press). Hepatica henryi might therefore be another, but autotetraploid, derivative of this species. Alternatively, and more likely, it could be an allotetraploid derivative of the entire-leaved H. asiatica as the putative paternal parent, and the crenate-leaved H. falconeri (or a related, extinct crenate-leaved species) as the putative maternal parent (assuming maternal inheritance of chloroplasts as in most angiosperms; Mogensen, 1996). Hepatica yamatutai is sometimes considered conspecific with H. henryi, but nuclear DNA content (Mabuchi et al., 2005) and molecular results clearly support its recognition as a separate species (Kokubun et al., 2004; this study). The origin of this species, however, is unclear. Plastid data firmly place it close to H. maxima, but the nuclear data are inconclusive in this respect. Finally, based on morphology and geographical position, H. transsilvanica could have originated from the European H. nobilis and the central Asian H. falconeri or its extinct crenate-leaved relatives. This hypothesis is clearly supported by the molecular data with some indication for H. nobilis var. nobilis as the maternal parent and H. falconeri as the paternal parent (Fig. 3). All hypotheses discussed earlier as to the origin of the tetraploids in the genus Hepatica are based on molecular and geographical evidence, and they should also be tested using the morphology and chromosome pairing behaviour of F1 hybrids (2x, 3x) obtained from crosses between H. falconeri and other putative parental diploid taxa or tetraploids themselves.

Genomic and chromosomal changes are often associated with hybridization and chromosome doubling, that is, allopolyploidy (Wendel, 2000; Otto, 2003). As exemplified here, such changes are not restricted to allopolyploidy, but also occur in autopolyploids (Weiss & Maluszynska, 2000), suggesting that chromosome doubling alone is sufficient to induce genome rearrangements. While allopolyploidy is now widely accepted as a major force, especially in plant evolution (Mable, 2003; Ma & Gustafson, 2005), researchers have just begun to appreciate autopolyploidy in a similar way, not least because of mounting evidence of its unanticipatedly high occurrence (Ramsey & Schemske, 1998, 2002; Suda, 2003; Stuessy et al., 2004; Yang et al., 2006). It is therefore expected that more detailed investigations will unravel many more examples of genome rearrangements in natural autopolyploids.


The authors’ appreciation is expressed to the Austrian National Science Foundation (FWF) for financial support under grant number P14825 to TFS; and to the FWF Hertha Firnberg Postdoctoral Fellowship program for financial support of HWS (project number T218). The authors also thank Dr H. Kato (Japan) and T.-K. Paek (Republic of Korea) for help with collecting the material.