Auto-and allopolyploidy in Centaurea sect . Acrocentron s . l . ( Asteraceae , Cardueae ) : karyotype and fluorochrome banding pattern analyses

Dysploidy and polyploidy are well documented in the large genus Centaurea, especially in sect. Acrocentron and in a small group of species from the Iberian Peninsula described as sect. Chamaecyanus, closely related to Acrocentron. We have explored two interesting cases of polyploid series in both sections: the polyploid series of Centaurea toletana in sect. Chamaecyanus and the series of C. ornata group in sect. Acrocentron. We have carried out a karyological study using both classic karyotype analyses and chromosome banding with fluorochromes. The C. toletana complex exhibits three different ploidy levels, diploid, tetraploid and hexaploid, the latter usually considered a different species, C. argecillensis. Our results do not contradict the original hypothesis of tetraploid C. toletana being an autopolyploid, but they suggest that this case of autopolyploidy, as many others, is the result of multiple events of polyploidization. Regarding C. argecillensis, our results confirm that it is an allopolyploid originated by hybridization of C. toletana and an unknown species. The colonizing habit of C. argecillensis, as it was the case of tetraploid C. toletana, suggests that the allopolyploid has multiple origins. The C. ornata complex exhibits four ploidy levels, diploid, tetraploid, hexaploid and endecaploid. In this section, however, the banding patterns have not helped us to determine the relation between the polyploid series and neither have allowed us to confirm the hypothesis of the possible progenitors of C. kunkelii.


INTRODUCTION
Until recently, polyploidy has been considered as an uncommon and isolated fact and polyploids have been thought genetically uniform, with low adaptive capacity and tending to disappear (STEBBINS, 1971).Currently, polyploidy is regarded as a prominent force in plant evolution (SOLTIS & SOLTIS, 2000;WENDEL, 2000;SOLTIS et al., 2003SOLTIS et al., , 2004).In the last 20 years, molecular studies in plants have changed the understanding of polyploidy evolution.The estimated frequency of polyploidy has increased, and it is now recognised that multiple origins are the rule of most polyploids (SOLTIS et al., 2003).
Traditionally, polyploidy has been thought to result from either duplication of a single genome (autopolyploidy) or from the combination of two or more differentiated genomes (allopolyploidy) (STEBBINS, 1947(STEBBINS, , 1950;;GRANT, 1981).However, GRANT (1981) also noted that both concepts are the extremes of a graded series.Autopolyploidy has historically been considered maladaptive and relatively uncommon compared to allopolyploidy (WENDEL & DOYLE, 2005).Whereas it is clear now that allopolyploidy is probably more common than autopolyploidy in nature, the latter is far more prevalent than was once thought.In addition, it is necessary to say that the autopolyploids never are maladaptive (RAMSEY & SCHEMSKE, 1998;WENDEL & DOYLE, 2005;SOLTIS et al., 2003SOLTIS et al., , 2004)).
An interesting case of study of polyploids is found in the genus Centaurea.Karyological features of Centaurea include both dysploidy and polyploidy.As to dysploidy, chromosome numbers in Centaurea range from x = 12 to x = 7 in a series well-correlated to phylogeny (GARCIA-JACAS et al., 2001).As to polyploidy, it is especially frequent in sects.Acrocentron (Cass.)DC. and Chamaecyanus Willk.(GARCIA-JACAS, 1998).
Centaurea sect.Chamaecyanus was described on the basis of a small group of species from the Iberian Peninsula (WILLKOMM & LANGE, 1870).Chamaecyanus is related to sect.Acrocentron, one of the largest of Centaurea (FONT et al., 2002), with about 100 species and a high percentage of endemics (GARDOU, 1975;WAGENITZ, 1975).The Ibero-North African region stands as the second most important centre of diversification after Turkey and the Balkans (GARCIA-JACAS & SUSANNA, 1992).Differ-ences between two sections rely only on the combination of acaulescent habit and reduced pappus in Chamaecyanus.Even though the boundaries are clear-cut, both sections are extremely close as demonstrated by the many fertile hybrids described between species of both sections (FERNÁNDEZ CASAS & SUSANNA, 1986) and subsectional rank looks certainly more appropriate for Chamaecyanus.Anyway, no matter the rank adopted (section or subsection), species of Chamaecyanus are probably a natural, monophyletic group, however unconfirmed by DNA sequence analyses (FONT et al., 2002).
The most extreme and interesting case of polyploidy in Chamaecyanus is C. toletana Boiss.& Reuter, which exhibits three different ploidy levels (2n = 2x = 20, 2n = 4x = 40, and 2n = 6x = 60).Tetraploid C. toletana was described as a different species, C. cavanillesiana Graells.However, careful examination of the type and new collections from the type localities in central Spain demonstrated that morphological differences between C. toletana and C. cavanillesiana were inexistent (FERNÁNDEZ CASAS & SUSANNA, 1986).In contrast, it is possible to characterize morphologically the hexaploid, which was also described as a different species, C. argecillensis Gredilla, later merged into C. toletana as a variety because many intermediate forms between the diploid type variety and the hexaploid exist (FERNÁNDEZ CASAS & SUSANNA, 1986).These authors concluded that the tetraploid was a true autopolyploid because it was impossible to differentiate the diploid from the tetraploid without carefully measuring pollen size or stigmas (LEWIS, 1980;RAMSEY & SCHENKE, 1998).However, FERNÁNDEZ CASAS & SUSANNA (1986) did not reach any conclusion on the hexaploid.
There is another case of similar polyploid series in sect.Acrocentron formed by three species C. gabrielis-blancae Fern.Casas (C.ornata var.microcephala Willk.), a diploid species, C. ornata Willd.a tetraploid and C. saxicola Lag. a hexaploid.DOSTÁL (1976) subordinated C. saxicola to C. ornata but a phylogenetic analysis of sect.Acrocentron (FONT et al., 2002) supports its status as a different species.In addition C. kunkelii, described recently by GARCIA-JACAS (1998) on the basis of a plant from south Spain that had been much confused with C. ornata, is also member of this group.Curiously GARCIA-JACAS (1998) suspected a hybridogenic origin for this endecaploid species, and one of the hypothesized parental species was C. saxicola, the other putative progenitor belonging to sect.Chamaecyanus, C. haenseleri a tetraploid species (FERNÁNDEZ CASAS & SUSANNA, 1986).
Interest for polyploidy has increased lately with the emergence of new molecular techniques (extensively revised in SOLTIS et al., 2004).Traditional views on meiotic pairing cannot be considered a valid criterion to distinguish between auto-and allopolyploids (QU et al., 1998).Instead, karyotype analysis may be very helpful in establishing systematic and evolutionary relationships within polyploid complexes, particularly those data which constitute what SCHWEIZER & EHRENDORFER (1983) named banding style, i.e. the distribution pattern of different chromosome regions, such as constitutive heterochromatin (revealed with Giemsa C-banding method) or AT-and GC-rich DNA portions (revealed with fluorochrome banding methods).We have used banding methodology a) to verify whether autopolyploidy of C. toletana is reflected in the banding pattern, b) to explore the relationships of hexaploid C. argecillensis to the diploid and tetraploid levels of C. toletana, c) to establish the relationship between C. gabrielis-blancae, C. ornata and C. saxicola and verify if they constitute a polyploid series and d) to look after a confirmation of the hypothesis that C. saxicola and C. haenseleri are the progenitors of C. kunkelii.

MATERIALS AND METHODS
Adult plants were collected from wild populations of each taxon.The species studied are listed in Table 1, with indication of their origins.
Chromosome counts were made on somatic metaphases using the squash technique.Root meristems from germinating seeds collected in the wild or from wild plants cultivated in pots were used.
Samples were pretreated with 0.05% aqueous colchicine at room temperature for 3 h as a minimum.The material was fixed with Carnoy for 24 h at -20ºC.Before staining, the material was hydrolyzed with 5N HCl for 1 h at room temperature.It was stained with 1% acetic orcein and mounted in 45% acetic acid.For all the counts, a minimum of ten plates (diploid and tetraploid taxa) or five plates (hexaploid taxa) from different individuals was examined.Preparations were made permanent by CO 2 freezing, dehydrating in ethanol and mounting in Canada balsam.The majority of photographs were taken using a Zeiss Standard microscope, and digital photographs were also taken using an Olympus 3030 camera mounted on an Olympus microscope U-TV1 X.The preparations, the negatives and the herbarium vouchers are preserved in the Botanical Institute of Barcelona (BC).Numerical parameters of the karyotypes were calculated and the idiograms were obtained with software running on Lotus 1-2-3 prepared by J. Simon (Universitat de Barcelona).
For fluorochrome banding, root tips were hydrolysed with 45% acetic acid for 5-10 min at 60ºC, and squashed in a drop of 45% acetic acid.After metaphase plate selection and cover slip removal by CO 2 freezing, slides were rinsed with absolute ethanol and dried at room temperature in covered containers until use.Two different fluorochrome banding protocols were applied to reveal GC-rich (with chromomycin A3, abridged CMA) and AT-rich (with bisbenzimide Hoechst 33258, abridged Hoechst) regions.For chromomycin A3 staining the techniques of SCHWEIZER (1976), KONDO & HIZUME (1982), COULAUD et al. (1995), CERBAH et al. (1995)

RESULTS AND DISCUSSION
A summary of cytogenetic results, including chromosome number, chromosomal formula, karyotype symmetry data and the number of GC-and AT-rich regions, is presented in Table 2, except for the endecaploid species C. kunkelii, for which it was only possible to obtain a single metaphasic plate and we have not been able to make the calculations.In this table we can appreciate the different chromosome sizes between the species belonging to sect.Acrocentron and those from sect.

Chamaecyanus.
Figure 1 (C.toletana series) and Fig. 2 (C.ornata series) show orcein-stained metaphase plates and chromosome morphology of the taxa studied.Karyotypes are very symmetrical in all the populations, and most chromosomes are metacentric, with some submetacentric ones present.All the species have satellites, not always in the same position.
In all the taxa, GC-and AT-rich zones were detected.Fluorochrome banded metaphase plates are shown in Fig. 3.The haploid idiograms of each population with the location of GC-and AT-rich regions are presented in Fig. 4. For the hexaploid population (C.argecillensis), the exact assignation of chromomycin and bisbenzimide bands in the idiogram was not performed (Fig. 4C), because of the extreme similarity in length and symmetry of all the chromosomes.We have not included either the idiograms of C. saxicola and C. kunkelii, since it was impossible to assign a banding patterns adjusted to the reality in these species.
Chromomycin-positive bands are often coincidental with bisbenzimide-negative zones.
In diploid C. toletana three telomeric bands are present in the chromomycin staining, while there is only one in the bisbenzimide staining, together with two negative satellites (Figs.3A, 3D, 4A).

Centaurea toletana complex
In C. toletana series, GC-and AT-rich regions are predominantly terminal.This banding pattern agrees with those of other Cardueae, such as that of the Xeranthemum group (GARNATJE et al., 2004), relatively close to the genus considered here.
In the tetraploid population, there are also three telomeric bands in the chromomycin staining with two more signals in two satellites (Figs.3B, 4B).In the bisbenzimide staining, however, there are two positive signals, one telomeric and another one in a satellite, together with a negative satellite (Figs.3E, 4B).Regarding C. argecillensis, we have already pointed out the impossibility of determining exactly the chromosomes where the signals are located, but there are five chromomycinpositive signals, three of them telomeric, one centromeric and the last one in a satellite (Fig. 3C).As for bisbenzimide, there are two telomeric bands and two negative satellites (Fig. 3F).
According to these results, the fluorescent banding pattern of the tetraploid population is not a perfect duplicate of the diploid one: even though diploid and tetraploid C. toletana studied populations share at least three of the signals, there are obvious differences between an ideal duplication of the 2x pattern and our results.
In our opinion, this does not contradict the hypothesis of autopolyploidy, which, as we have seen, has very sound basis.In fact, we could not expect the tetraploid to be a perfect duplication of the genome of the diploid, which would be possible only in an ideal case that we would seldom find in nature.Many factors are responsible for these differences, but we can stress upon two obvious ones.
First factor is relative age of polyploidy.If tetraploid populations have long evolved without ge-    SUSANNA, 1986).Second, and probably the most important factor, tetraploid C. toletana is a typical colonizing polyploid (STEBBINS, 1950;LEVIN, 1983).While diploid populations of C. toletana grow only on siliceous soils, tetraploid populations have a wider ecological tolerance and have colonized a very large extension in central Spain on clay soils, occupying frequently somewhat disturbed habitats (FERNÁNDEZ CASAS & SUSANNA, 1986).Multiple origins of these successful, colonizing polyploids are now accepted as a rule (SOLTIS &SOLTIS, 1993 and1995;VAN DIJK & BAKX-SCHOTMAN, 1997).The model of distribution of diploid C. toletana also supports a multiple origin of the tetraploid race: the diploid cytotype grows in small and disjunct populations in which genetic drift could easily lead to partial reproductive isolation, favouring the occurrence of polyploidy, as was suggested for Centaurea and related genera by VILATERSANA et al. (2000) and GARNATJE et al. (2001).Our preliminary tests using RAPDs (FONT, unpublished data) suggest that there are differences between diploid populations of C. toletana.Thus, the origin of the tetraploid would be in multiple events of intraspecific hybridization between partly genetically isolated diploid individuals, rather than a single duplication in a population (intraspecific hybridization has been suggested for many other autoployploids by SOLTIS et al. (2004).This would justify the differences found between a simple duplication of the 2x genome and our results.
Regarding hexaploid C. argecillensis, we have found no cytogenetic connections to the two other cytotypes of the polyploid series.An outstanding difference is the presence in the hexaploid species of a centromeric chromomycin-positive region, absent from both ploidy levels (diploid and tetraploid) of C. toletana, making it difficult to point out autopolyploidy as the mechanism of origin for the 6x taxon from the 2x and 4x one.Morphological relationships to C. toletana are so obvious that cannot be disregarded, and populations of C. toletana with tetraploid cytotype are common in the area of distribution of C. argecillensis.However, C. argecillensis is very probably an allopolyploid: the extreme morphological variability pointed out by FERNÁNDEZ CASAS & SUSANNA (1986) also suggests a hybrid origin.As was the case for the tetraploid, a multiple origin of the hexaploid is very probable, not only because of its variability, but also because the species is a very active colonizer (FERNÁNDEZ CASAS & SUSANNA, 1986).The question that remains open is the identity of the other species involved in the origin of C. argecillensis.Achene morphology is very reliable here: the wanting parental species must be probably a species of sect.Chamaecyanus because hybridization with species of sect.Acrocentron is reflected in the length of the pappus and in the frequent loss of the caulescent habit (revised in depth by GARCIA-JACAS, 1998).The identity of the second species that has originated C. argecillensis is a question that further studies using molecular markers and fluorescent and genomic in situ hybridization will try to make clear.

Centaurea ornata complex
Regarding the species belonging to the C. ornata series, we can observe in the diploid C. gabrielisblancae the presence of four markings with the chromomycin staining, a positive centromeric band and another telomeric, and the presence of two negative satellites.However, we have found a single centromeric marking in the larger chromosome in the case of the bisbenzimide staining (Figs. 3G,  4D).In the two tetraploid species we observe four positive markings, three telomeric bands and a centromeric one, and a negative satellite marking with CMA in C. ornata, while with Hoechst staining we can see three telomeric markings.In C. haenseleri we also observe four positive CMA bands, three telomeric and one centromeric, together with a negative satellite.We must recall that the exact position of marks in two tetraploid species is different (Figs.3H, 3I, 4E, 4F).
With respect to the hexaploid and endecaploid species, we can only point out the number of signals observed, since the high similarity in symmetry of the whole karyotype does not allow us to check which chromosomes bear the marks.In C. saxicola we can see with CMA staining a minimum of 10 markings, and also 10 with Hoechst, of which some are centromeric, and the presence of some negative satellites is also shown.
In C. kunkelii we 16 to 18 CMA markings and a minimum of six Hoechst bands, with two positive satellites.
In accordance with the results obtained, species from the C. ornata complex do not present a banding pattern allowing us to determine which has been the origin of the polyploid series.The bands present in the diploid species do not coincide with those revealed in the tetraploid and hexaploid species.A possible explanation to this fact could be the intra-and intergenomic reorganization of polyploid genomes, which can be extensive and may occur rapidly (SOLTIS et al., 2003).In addition, tetraploid C. ornata does not share the same habitat that C. gabrielisblancae (a diploid species) and C. saxicola (hexaploid).We must hypothesize that the formation of C. saxicola is an ancient event, and the long time ellapsed allowed extensive genomic reorganization.
The banding patterns obtained make it difficult to point out any relationship with C. kunkelii (endecaploid species); since it is not possible to determine to which chromosomes the bands belong.Anyway, the number of bands found in C. kunkelii does not agree with the number obtained in the other species.Centaurea kunkelii is known from a single population, which probably implies a long isolation.This isolation could also be the reason for a deep restructuring of the genome that has evolved in a different way.
In later analyses using nuclear-ribosomal DNA sequences of the ETS region (FONT et al., unpublished data) we have verified that the diploid and tetraploid species are grouped in the Iberian clade, whereas C. saxicola and C. kunkelii belong to a clade with the African species and some Iberian ones like C. haenseleri, C. prolongi Boiss.and C. lainzii Fern.Casas with which they share the area of distribution in South Spain.These results do not contradict our starting hypothesis that the possible progenitors of C. kunkelii are C. saxicola and C. haenseleri.Conversely, the ETS-based phylogeny groups C. gabrielisblancae and C. ornata, a relationship that does not agree with the very different banding patterns obtained in the present work.In the case of C. ornata, this lack of correlation probably is caused by the multiple origins of polyploid species as previously stated.

Table 1 .
Origin of the studied populations.