|Home | About | Journals | Submit | Contact Us | Français|
Studying the spatial distribution of cytotypes and genome size in plants can provide valuable information about the evolution of polyploid complexes. Here, the spatial distribution of cytological races and the amount of DNA in Dianthus broteri, an Iberian carnation with several ploidy levels, is investigated.
Sample chromosome counts and flow cytometry (using propidium iodide) were used to determine overall genome size (2C value) and ploidy level in 244 individuals of 25 populations. Both fresh and dried samples were investigated. Differences in 2C and 1Cx values among ploidy levels within biogeographical provinces were tested using ANOVA. Geographical correlations of genome size were also explored.
Extensive variation in chromosomes numbers (2n = 2x = 30, 2n = 4x = 60, 2n = 6x = 90 and 2n = 12x =180) was detected, and the dodecaploid cytotype is reported for the first time in this genus. As regards cytotype distribution, six populations were diploid, 11 were tetraploid, three were hexaploid and five were dodecaploid. Except for one diploid population containing some triploid plants (2n = 45), the remaining populations showed a single cytotype. Diploids appeared in two disjunct areas (south-east and south-west), and so did tetraploids (although with a considerably wider geographic range). Dehydrated leaf samples provided reliable measurements of DNA content. Genome size varied significantly among some cytotypes, and also extensively within diploid (up to 1·17-fold) and tetraploid (1·22-fold) populations. Nevertheless, variations were not straightforwardly congruent with ecology and geographical distribution.
Dianthus broteri shows the highest diversity of cytotypes known to date in the genus Dianthus. Moreover, some cytotypes present remarkable internal genome size variation. The evolution of the complex is discussed in terms of autopolyploidy, with primary and secondary contact zones.
Polyploidy, defined as the possession of at least three complete sets of chromosomes, is a widespread phenomenon in the flowering plants. It occurs in 30–35 % of the angiosperms according to Stebbins (1971), whereas Coghlan et al. (2005) estimated that 50–70 % of all species could be polyploid. Polyploidy has significant effects on the biochemistry, ecophysiology and morphology of plants (Levin, 1983; Thompson et al., 1997; Levin, 2002; Pearse et al., 2006; Buggs and Pannell, 2007), but the most obvious effect is increasing overall DNA amount following doubling of chromosome sets.
Knowledge on the DNA content of the non-replicated monoploid genome (i.e. the 1Cx value sensu Greilhuber et al., 2005) may be useful in comparisons across polyploid complexes (Smarda et al., 2008a). It seems monoploid genome size can respond to polyploidization in different ways, as some studies show decreased 1Cx values following polyploidization, others document enlarged values (Ozkan et al., 2003; Leitch and Bennett, 2004; Bennetzen et al., 2005), and still others report constant 1Cx values within polyploid series (Ohri, 1998; Ozkan et al., 2006). Therefore, detecting patterns of change in 1Cx values within a given taxonomic group may allow for inferences about phylogenetic history.
Cytological races within polyploid species can be spatially/ecologically segregated, which is explained on the basis of minority cytotype exclusion (Levin, 1975) or varying ecological tolerances (Van Dijk and Bakx-Schotman, 1997; Mandakova and Munzbergova, 2006). Alternatively, several cytotypes can coexist within a population, and this is often interpreted as suggestive of a variety of evolutionary/ecological process that operate simultaneously (e.g. Heuchera grossulariifolia, Segraves et al., 1999; Ranunculus adoneus, Baack, 2004; Solidago altissima, Halverson et al., 2008). It follows that knowledge on geographical distribution of cytotypes and their ecological correlates can be useful to give insights on speciation.
Carolin (1957) studied ploidy levels in 91 species of the genus Dianthus and found most (67 %) were diploid, even though polyploidy was not uncommon: tetraploid species amounted to 18·7 %, hexaploids to 6·6 %, and the remaining (7·7 %) presented several cytotypes. In the section Plumaria of Dianthus to which D. broteri belongs, a later study by Weiss et al. (2002) showed a majority of species containing two or three ploidy levels. Tetraploids and hexaploids were most frequent in the section, whereas pentaploids, triploids and diploids were relatively rare. Furthermore, 2x cytotypes appeared mainly in isolated glacial refugia and were less common than 4x and 6x. Weiss et al. (2002) also reported a few cases of populations composed of cytotype mixtures.
Dianthus broteri is a subshrub with large, fragrant flowers pollinated by crepuscular hawkmoths (F. Balao, unpubl. res.). Endemic to the Iberian Peninsula, it invariably lives in locations close to the coast from the south of Portugal to the east of Spain (Fig. 1). Although some authors (Boissier and Reuter, 1852; Lemperg, 1936; Bernal et al., 1990) considered D. broteri a close relative of North African D. serrulatus Desf., recent sequencing studies of the two taxa using ITS and cpDNA (P. Vargas et al., RJB, Madrid, Spain, unpubl. res.) indicate that they are only distantly related. In D. broteri both tetraploid and octoploid populations are known (Carolin, 1957; Coy et al., 1997), and the latter have been described by Gallego (1986) as a ‘splinter’ species (D. inoxianus Gallego, with 2n = 120).
In the present study, an analysis of the cytotypes of D. broteri s.l. is carried out using chromosome counts and flow cytometry. The latter technique is more and more used in studies of genome size and polyploidy, its main advantage being that it allows the analysis of large numbers of randomly chosen individuals (Kron et al., 2007). Furthermore, it can be performed on dehydrated plant material in some species (although care must be used in applications other than ploidy determination; Suda and Travnicek, 2006). Many recent reports support flow cytometry as a valid technique to assess intraspecific genome size variations (e.g. Pecinka et al., 2006; Walker et al., 2006; Smarda et al., 2008b). The main goals were: (a) to ascertain which and how many levels of ploidy exist, as well as their geographical distribution; (b) to estimate genome size and investigate any possible relationships with environmental parameters; (c) to make reasonable inferences on the evolutionary history of polyploidy in this complex, based on monoploid genome size and geographic distribution data.
During the summers of 2006 and 2007, 244 individuals chosen at random were sampled in 25 populations of Dianthus broteri Boiss. & Reut. distributed around the distribution range of the species (Fig. 1). This range covers four biogeographical provinces (Rivas-Martínez et al., 2004) which present different types of the Mediterranean climate. For example, the Coastal Lusitan-Andalusian province is markedly influenced by the Atlantic Ocean and as a result enjoys relatively benign ecological conditions with regard to rainfall and temperature (no winter frosts). The adjoining Betic province is bioclimatically heterogeneous and composed of plains and mountain ranges bordering the Guadalquivir River. With dry to perhumid regimes, biodiversity is accordingly high in this region, which also has high rates of paleoendemism. East of the Betic province, the Murcian-Almeriensian region presents subarid to arid conditions and relatively cold winters. Lastly, the Balearic-Catalan-Provençal province (partly insular) has a coastal variety of the Mediterranean climate and rugged topography. All four provinces share high rates of human disturbance (e.g. wildfires are common), although the intensity of perturbations tends to decrease from lowlands to mountains.
Precise details of the localities and populations are provided in the Table 1. At each site, seeds and leaves were collected from the plants and preserved in individual, numbered containers. Leaves were then stored in silica gel, and seeds sown in individual pots in the greenhouse of the University of Seville. Voucher specimens for all populations are deposited at the Herbarium of the University of Seville (SEV, Spain).
Seeds from 59 plants of nine populations (Table 1) were germinated on moist filter paper in a germination chamber at 25 °C, under a 16 : 8 light : darkness regime. Root tips 1–3 cm long were collected and treated with 0·002 m 8-hydroxyquinoline for 3 h at 4 °C, then fixed in 3 : 1 ethanol : acetic acid at 4 °C for a minimum of 24 h. Following staining with alcoholic hydrochloric acid-carmine (Snow, 1963) at 30 °C for 3 d, root tips were squashed on a microscope slide. Photomicrographs of chromosomes at mitotic metaphase were taken with a ×63 Carl-Zeiss Axiophot photomicroscope (Carl-Zeiss Inc., Germany) equipped with Sony DXC-390P Exwave HAD (Tokyo, Japan). Snapshots were exported and studied using image analyser software.
The absolute nuclear DNA content of 244 individuals, (6)8–10(19) per population, was determined by flow cytometry (FCM hereafter) between February and May of 2008. Most often fresh leaf material taken from greenhouse-grown seedlings was used, but for a few populations in which there was no alternative (see Table 1) silica-dried leaves collected in the field were employed.
Nuclear suspensions for FCM were prepared following the protocol of Dolezel et al. (1989). Approximately 100 mg of leaf tissue were chopped with a razor blade onto a Petri dish containing 1 mL of LB01 buffer. The resulting suspension was ﬁltered through 30-μm mesh CellTric disposable ﬁlter (Partec GmbH, Münster, Germany), and stained with 50 µg mL−1 of propidium iodide and 50 µg mL−1 RNase. Until analysis, samples were kept on ice and protected from light. FCM measurements were taken at facilities of the Biology Research Services (CITIUS, University of Seville), using a Coulter CYTOMICS FC500-MPL (Beckman Coulter, Fullerton, CA, USA) equipped with a 20-mW argon-ion laser at 488 nm.
The primary internal standard used to determine DNA amount was Pisum sativum ‘Ctirad’, a taxon with 9·09 pg of 2C nuclear DNA. For dodecaploid material, and because peaks overlapped, a tetraploid specimen of D. broteri was used as a secondary standard which was calibrated against P. sativum. The genome size for this tetraploid specimen was 2C = 3·46 pg (nine replicate measurements performed on different days). Peak means were established through manual gating using CXP software (Beckman Coulter).
Overall, 732 FCM measurements were performed, with three replicates per plant. Replicate measurements were taken on different days in order to minimize instrumental drift. Except for six readings from five individual plants which were discarded due to poor sample quality (i.e. <1000 nuclei per peak), all measurements were used for analyses (i.e. within-plant extreme values were not dismissed). Absolute DNA content was calculated for each sample as the average of the three replicates, and monoploid genome size (1Cx) estimated as the amount of nuclear DNA divided by ploidy level. A few samples with genome size variation greater than 5·5 % (either within a population, or within a given ploidy level) were run simultaneously to confirm that the divergences were not due to methodological artefacts.
Differences in ploidy levels among populations and biogeographical provinces were evaluated using analysis of variance (ANOVA) and Tukey's HSD post-hoc tests. Dependent variables for the analyses were absolute DNA content (2C value) and monoploid genome size (1Cx). Relationships between relative genome size (population means) and geographic variables (longitude, latitude and altitude) were investigated using Spearman's rank correlations. A Mantel test based in Spearman's rank correlation with 1000 permutations was used to test the relationship between geographical distance and monoploid genome size. All analyses were performed with R statistical software (R Development Core Team, 2008).
Photomicrographs of chromosome complements are presented in Fig. 2. Each population had a single level of ploidy, which could range between 2n = 30 (pop. 3; Fig. 2A), 2n = 60 (pops 1, 11, 12, 23; Fig. 2B), 2n = 90 (pop. 18; Fig. 2C) and 2n = 180 (pop. 8; Fig. 2D). As the basic number in the genus Dianthus is x = 15 (Carolin, 1957), these chromosome numbers would correspond, respectively, to diploid, tetraploid, hexaploid, and dodecaploid levels. Chromosomes were of small size (1–2·5 µm), metacentric or submetacentric. Satellited chromosomes were observed in all of the cytotypes observed.
Genome size and ploidy levels in 25 populations of Dianthus broteri are shown in Table 1. In general, all ploidy levels showed clear DNA peaks (Fig. 3), with observed ratios of 1·00 : 1·56 : 2·06 : 2·88 : 5·58, respectively, for 2x, 3x, 4x, 6x and 12x cytotypes. FCM measurements revealed that most nuclei were in G0/G1 phase, although in young leaf samples (i.e. growing very actively) a considerable proportion of nuclei in phase G2 existed. As the DNA content of G2 nuclei invariably doubled that of the G0/G1 phase, it seems safe to assume that measurements of DNA content were linear.
Measurement error (calculated as the standard deviation of three replicates per individual) was smaller than 0·015 in 95 % of cases, so this value was assumed to represent the true deviation of individual measurements at P = 0·05. A negative correlation existed between the sample coefficient of variation (CV) and ploidy level (Spearman's γs = –0·79, P < 0·001), but this was likely to be a methodological artefact, as samples were always run with the same instrument settings (see Discussion).
The CV for peaks varied between 1·4 % and 4·5 % (mean ± standard deviation, 3·64 ± 1·35 %) and, in general, fresh material yielded significantly better readings (average CV = 2·8 %) than dehydrated samples (CV = 4·5 %; Mann–Whitney U = 1727, P < 0·001). Nevertheless, drying did not modify DNA amount estimations in any important ways: in a subset of nine populations including dried and fresh samples that were analysed with ANOVA, the 1Cx value was independent of tissue state (F1,68 = 1·263, n.s.; for the population effect, F8,68 = 33·150, P < 0·001; note, however, that the interaction term could not be tested in this case). For the only population which had both dried and fresh leaves analysed (pop. 3), means were statistically indistinguishable (two-samples t test, t = 0·605, d.f. = 12, n.s.).
The geographic distribution of cytotypes as estimated from FCM is shown in Fig. 1. Two disjunct areas with diploid populations were apparent, one in south Portugal (pops 3 and 4) and another along the Betic Ranges in S Spain (pops 13–16). Tetraploid cytotypes were widespread (11 populations, or 44 % of total) and occupied two disjunct areas in east and south-west Iberia. In contrast, hexaploid and dodecaploid cytotypes were relatively rare and spatially restricted, the former to an area east of the Betic Ranges (pops 17–19) and the latter to the lower Guadalquivir River valley (pops 5–9). Except for one population with diploid and triploid plants (pop. 3), the remaining populations showed a single cytotype.
As expected, ploidy level was a major determinant of DNA amount (Fig. 4A; ANOVA: F4,239 = 22105; P < 0·001), and the mean 2C value was significantly different among any two cytotypes. For any ploidy level, the DNA content varied more among than within populations (ANOVA: F25,218 = 104·9; P < 0·001).
As listed in Table 1, the amount of 2C DNA varied 5·86-fold, from 1·70 pg in diploids (2n = 30) to 9·96 pg in dodecaploids (2n = 180). Yet this variation was not solely due to ploidy level, as 2C DNA amount could also vary to some extent within cytotypes. For diploid plants, for example, the 2C value varied between 1·70 and 1·99 pg (i.e. a 1·17-fold variation); triploids ranged from 2·86 to 2·90 pg (about 1·03-fold); tetraploids from 3·07 to 3·74 pg (1·22-fold); hexaploids from 5·00 pg to 5·61 pg (1·12-fold); and dodecaploids from 9·20 to 9·96 pg (1·08-fold).
The monoploid DNA amount (1Cx) ranged between 0·78 pg and 1·00 pg and, according to ANOVA, was significantly dependent on ploidy level (F4,239 = 81·65, P < 0·001). Post-hoc comparison tests failed to reveal any significant differences in 1Cx value between tetraploids and hexaploids (Fig. 4B). In general, the amount of monoploid DNA was inversely related to ploidy level (i.e. 2x, 3x, 4x, 6x and 12x cytotypes, respectively, had mean 1Cx values of 0·91, 0·96, 0·86, 0·87 and 0·81 pg).
Variation in genome size was usually <5·5 % within populations. In the few examples with higher variability (up to 8·7 % in populations 3, 5, 12 and 24), simultaneous FCM analyses of the samples confirmed that these unusual values had been caused by methodological artefacts. Additionally, concurrent runs demonstrated substantial among-population differences (within diploids and tetraploids) which resulted in distinct fluorescence peaks (Fig. 5).
In general, monoploid DNA content was unrelated to altitude, latitude or longitude. Correlations between elevation and monoploid DNA amount were largely non-significant, although significant relationships were detected (for tetraploids only) with latitude (r = –0·89, n = 11) and longitude (r = –0·69, n = 11). This implied increasing 1Cx values from north to south and from east to west. A Mantel test failed to detect any broad spatial autocorrelations for monoploid DNA amount (γM = 0·10, n.s.), although some structuring existed at smaller spatial scales (e.g. populations within 100 km of each other; Mantel correlogram γM = 0·22, P = 0·002).
Combining cytotypes and biogeographic provinces resulted in eight groups (presented in Fig. 6) that were subject to ANOVA for an assessment of monoploid genome size variation (F7,237 = 121·85, P < 0·001). According to post-hoc tests, Coastal Lusitan-Andalusian diploids and triploids were indistinguishable as regards monoploid DNA (overall mean 0·95 ± 0·03 pg), and so were Coastal Lusitan-Andalusian tetraploids, Betic diploids and tetraploids, as well as Murcian-Almeriensian hexaploids (0·89 ± 0·03 pg). Balearic-Catalan-Provençal tetraploids (0·83 ± 0·03 pg) and dodecaploids (0·81 ± 0·02 pg) presented 1Cx values that were in both cases significantly smaller than those of the remaining populations.
Many species in the subfamily Caryophylloideae have diploid and tetraploid cytotypes, although only Silene ciliata Pourret was known previously to present a long polyploid series (2x, 3x, 4x, 10x, 14x, 16x, 18x, 20x and 30x; Küpfer, 1974). The present data document the most extensive polyploid series known to date for Dianthus, with five ploidy levels (2x, 3x, 4x, 6x and 12x). The complexity of this series is comparable to those reported in Senecio carniolicus (Suda et al., 2007) and Claytonia perfoliata (Miller, 1976).
Dianthus broteri was previously considered a tetraploid taxon, and this was actually the most common cytotype in the present study (44 % of all populations). The relatively rare dodecaploid cytotype (2n = 180) was restricted to five populations west of the Guadalquivir Valley, and is described here for the first time. Gallego and Talavera (1993) reported a plant with 2n = 8x = 120 chromosomes at one of the dodecaploid sites (pop. 7), but since no octoploids were found at the study sites their count might be incorrect.
In general, the present study revealed considerable intraspecific variations in genome size, but these cannot be automatically assumed to be real. Greilhuber (2005) demonstrated that methodological artefacts can contribute to variability, so one must be cautious with interpretations. We believe within-cytotype variation reported in the present study for D. broteri (up to 1·22-fold in simultaneously run tetraploids; Fig. 5) should be considered genuine, as our samples were from young plants, grown in a common environment, replicated well (standard deviation = 0·015) and had low coefficients of variation (3·64 % on average, thus making any effect of secondary metabolites on measurements unlikely). In contrast, within-population variability (5·5 % on average) could not be established firmly through simultaneous runs. The likely reason behind was insufficient sensitivity of the method; as shown by Benson and Braylan (1994) and Dolezel and Göhde (1995), two individual peaks can be distinguished only if they differ in an amount that doubles their coefficients of variation. In the present study this means differences greater than 5·6 (CVs were around 2·8 at best in the present study), so within-population variability remained undemonstrated.
Another point that requires interpretation in the present data was the correlation between the coefficients of variation of samples and levels of ploidy (see Results). This was certainly a by-product of the fact that low ploidy levels were read (as it is often done) on lower channels (i.e. with lower voltage) of the flow cytometer. These channels are invariably less precise and, as instrumental settings were kept constant throughout, relatively higher internal variation could not be avoided in low polyploid samples.
The use of dry material to study ploidy levels is relatively infrequent, and has been subject to criticism. In their comparative study, Suda and Travnicek (2006) showed that results are to a great extent species-dependent, and concluded that fresh material should always be preferred over dried samples for studies of absolute genome size (mostly because drying decreases fluorescence and a parallel decline in measurement reliability; i.e. coefficients of variation get larger). However, in taxa which do not experience strong decreases in fluorescence (like D. broteri in the present study) measurements of dehydrated leaves can still be valid (see also D. gratianopolitanus; table 1 in Suda and Travnicek, 2006). If this condition is met, dried samples can be the most suitable material to study absolute genome size in large numbers of plants, or even herbarium vouchers. Suda and Travnicek (2006) used DAPI in their study, but the present data indicate that propidium iodide can also be used to obtain reliable measurements.
As regards spatial and ecological correlates, the effects of elevation and geographical location on genome size were in general weak. Size varied significantly with latitude/longitude only in the tetraploids, although not continuously (rather, the correlation was due to clustering in southern and eastern areas). Furthermore, biogeographical provinces did not show clear-cut differences as regards genome size (Fig. 6), making unlikely a role of soil conditions and/or climate (cf. Bituminaria bituminosa, Walker et al., 2006; Ceratonia siliqua, Bures et al., 2004; Festuca pallens, Smarda and Bures, 2006). Lastly, genome size followed the general trend (Leitch and Bennett, 2004) to become smaller with increasing ploidy levels (Fig. 4B), although differences were statistically significant in a few cytotypes only (for example, dodecaploids had the lowest 1Cx value whereas diploids showed the highest).
Information on genome size given in this study, combined with geographical distribution of cytotypes, can provide some insight on the origin and evolution of the polyploid series. In short, there were five ploidy levels (including triploids) which also present some internal variability as regard DNA amount. The most parsimonious explanation for the fact that diploid, tetraploid and hexaploid populations along the Betic Ranges, as well as tetraploid populations of the Atlantic coast were all alike in terms of monoploid DNA amount (Fig. 6) is that most of these cytotypes originated by autopolyploidy. Furthermore, the differently sized genomes of Coastal Lusitan-Andalusian diploids (i.e. Serra de Monchique and Serra do Caldeirão) vs. those of Betic areas (i.e. Sierra Nevada and Sierra Tejeda-Almijara) could be a symptom of prolonged separation among them. Confinement of diploid cytotypes to southern areas has also been observed in other Dianthus species (Weiss et al., 2002) and suggests these areas may have acted as refugia during Pleistocene glaciations. Actually, these two geographic areas have been recognized previously as refugia for other species including trees (e.g. Quercus spp., Olalde et al., 2002; Petit et al., 2002; Pinus sylvestris, Sinclair et al., 1999) and herbs (e.g. Senecio gallicus, Comes and Abbott, 1998; Armeria spp., Larena et al., 2006).
It is also noteworthy that southern and eastern tetraploid populations had significantly different 1Cx values (Fig. 6). This suggests they may have experienced different evolutionary histories and/or independent polyploidization events, a notion supported by results of experimental crosses (F. Balao; unpubl. res.) between 4x plants from the Betic Ranges (pop. 11) and eastern Spanish counterparts (pop. 25). In these crosses, only 42·8 % of ovules were transformed into mature seeds which either produced achlorotic cotyledons or failed completely to germinate. A solid reproductive barrier seems to exist between tetraploid cytotypes which, in addition, show a number of morphological differences: southern tetraploids have smaller flowers and less laciniated petals than eastern tetraploids (described in the past as a different species, D. valentinus; Willkomm, 1859). The different 1Cx values of eastern and southern tetraploids could also be due to introgression with other taxa, although this hypothesis has yet to be tested. Some authors (Bernal, 1989; Bernal et al., 1990) have suggested introgression between eastern D. broteri tetraploids and D. multiaffinis Pau, a species that also occurs in the Iberic Mountains. Introgression is not uncommon in Dianthus (Andersson-Kottö, 1931; Carolin, 1957; Nimura et al., 2003, 2008), and Gatt et al. (1998) reported that crossings of D. caryophyllus (2x) and D. plumarius (6x), or D. plumarius (6x) and D. knappi (2x) produce tetraploid hybrids.
As a rule, the populations studied contained a single cytotype. This may indicate an effect of minority cytotype exclusion (i.e. relatively rare cytotypes are prone to disappear in mixed populations; Levin, 1975) and/or a widespread lack of gene-flow between ploidy levels in nature. In Centaurea jacea (Hardy et al., 2001) and Chamerion angustifolium (Husband and Sabara, 2004) plants with different levels of ploidy are effectively isolated reproductively, and this may also apply to D. broteri.
The above rule had two exceptions, namely (1) neighbouring dodecaploid and tetraploid populations of the lower Guadalquivir valley, and (2) one population from Portugal (Monte Clérigo, pop. 3) which had a majority of diploids along with some triploid plants. These two cases of cytotype coexistence could be reasonably explained in terms of ‘primary’ or ‘secondary’ contact zones. According to Petit et al. (1999), a primary contact zone is an area where a cytological race gives rise to another, and then the two coexist sympatrically. This seems the case of the Portuguese population composed of a majority of diploids together with a few triploids which (we hypothesize) could have originated locally by fusion of (diploid) unreduced gametes. In accordance with this view the amounts of monoploid DNA of the coexisting cytotypes were identical.
In turn, neighbouring dodecaploid and tetraploid races in the lower Guadalquivir valley could be best explained in terms of a secondary contact zone (sensu Petit et al., 1999; i.e. neither of the cytotypes can be claimed to have acted as a parent of the other and, due to genetic differences, plants become relatively intersterile). Secondary contact zones have been described in Aster amellus agg. (Mandakova and Munzbergova, 2006), Knautia arvensis agg. (Kolar et al., 2009) and other polyploid complexes (Thompson and Lumaret, 1992). In the particular case of D. broteri, the dodecaploid and tetraploid races seem largely unrelated to each other since they differ as regards monoploid DNA amount (Fig. 4B). Furthermore, numerous ex situ and in situ attempts at crossing dodecaploid and tetraploid plants invariably failed (F. Balao; unpubl. res.), which suggests they are reproductively isolated. It is also remarkable that dodecaploid D. broteri have scabrid stems, a feature not found elsewhere [and used as a diagnostic character for D. inoxianus, the splinter species described by Gallego (1986)]. In summary, results of the present study indicate that there is a diversity of processes in the evolution of D. broteri cytotypes, but more research and molecular studies are still needed to get a clearer picture of the phylogeography and evolution of the complex.
We thank Dr J. Dolezel for supplying the standard Pisum seeds, and M. Carballo and A. García-Quintanilla for technical assistance with the cytometry. Dr P. Gibbs and two anonymous reviewers provided invaluable comments on a previous version of the manuscript. This study was supported by a predoctoral grant to F. Balao from Spanish Ministerio de Educación y Ciencia (AP2005-4314); Junta de Andalucía, Proyecto de Excelencia (2005/RNM848) and Ministerio de Educación y Ciencia, Flora Iberica 7(2)[CGL2006-00817]. The Doñana National Park (Patronato del Parque Nacional de Doñana), gave permission to access to several sampling sites (project 34/2004).