Endopolyploidy in seed plants is differently correlated to systematics, organ, life strategy and genome size


Correspondence: Armin Meister. Fax: 0049 039482 5137; e-mail: meister@ipk-gatersleben.de


Endopolyploidy is a systemic feature in seed plants. A negative correlation between genome size and endopolyploidization has been claimed previously, assuming that a minimum amount of DNA, necessary for certain cell functions, has to be acquired by endopolyploidization of the corresponding cells in plants with small genomes. This assumption was based on the analysis of only a limited set of data from few species. In the present study the endopolyploidization of several organs of 54 seed plant species belonging to two gymnosperm and 14 angiosperm families was investigated. The results revealed a low negative correlation between genome size and endopolyploidization. However, differences between the families, between the different organs of a given species and between the different life-cycle types with regard to endopolyploidization became obvious. A three-way analysis of variance with covariate to quantify the impact of the different factors on the extent of endopolyploidization suggested that taxonomic position is the major factor determining the degree of endopolyploidy within a species, while life cycle, genome size and organ type have a minor but also significant effect on endopolyploidization. The comparison of habitats of 16 investigated Central European species implies that endopolyploidization represents a mean to accelerate the growth of plant species in niches, which require and support fast development.


Endopolyploidization, i.e. the existence of different ploidy levels (labeled as 2C, 4C, 8C…) in adjacent cells of a species, is a common phenomenon in seed plants.

Nevertheless, the biological significance of endopolyploidy is not yet clear. Endopolyploidy has mainly been found in species with small genomes (e.g. Arabidopsis thaliana, Galbraith, Harkins & Knapp 1991; Cucumis sativus, Gilissen et al. 1993; Lycopersicon esculentum,Smulders, Rus-Kortekaas & Gilissen 1994; Medicago sativa and Medicago truncatula, Kondorosi, Roudier & Gendreau 2000; and Brassica oleracea, Kudo & Kimura 2001a, b). Among succulents, species with small genomes such as Mesembryanthemum crystallinum, but not those with large genomes revealed endopolyploidization (De Rocher et al. 1990). Accordingly, Nagl (1976) claimed a negative correlation between genome size and extent of endopolyploidization in animals and seed plants. A minimum amount of nuclear DNA in species with small genomes, frequently realized by endoreduplication, was assumed to be required to maintain the regulatory and functional state of certain cells (Nagl 1976; De Rocher et al. 1990; Galbraith et al. 1991). At least for some tissues this assumption seems to be unlikely because of the tight proportionality between ploidy level and cell volume reported for plant cells investigated with respect to this relationship (Bradley 1954; Melaragno, Mehrotra & Coleman 1993), since the gene dosage per cell volume remains constant over all ploidy levels. Furthermore, the wide range of ploidy levels within one cell type found by Bradley (1954) and Melaragno et al. (1993) indicates that no particular DNA amount is necessary to maintain the cell function. However, from the positive correlation between genome size and life cycle (short-lived species have small and perennial species large genomes; Bennett 1972; Vinogradov 2001) on the one hand and a negative correlation between genome size and endopolyploidization (Nagl 1976) on the other a third correlation would logically follow, namely a negative correlation between endopolyploidization and life cycle (preferentially species with short life cycles should show endopolyploidy).

Moreover, endopolyploidization behaviour seems to be typical for some seed plant families (Tschermak-Woess 1956; D’Amato 1964; Nagl 1976; Olszewska & Osiecka 1982).

To test the impact of genome size, taxonomic position and life cycle on endopolyploidization we investigated 54 seed plant species belonging to 16 families which were selected to cover a wide range of genome size within each family according to Bennett & Leitch (2001).

Up to 12 organs per species were analysed as in plants endopolyploidization seems to be typical for specific organs (De Rocher et al. 1990; Galbraith et al. 1991; Kudo & Kimura 2001a, b).

Finally, we included some ecological features of the tested Central European species, since Grime & Mowforth (1982) and Grime, Shacklock & Brand (1985) found a different growth rate for species with small and large genomes (which means small and large cells, respectively) in dependency on temperature due to different sensitivity of cell enlargement and cell division to low temperature. This different sensitivity to temperature may be a crucial point for endopolyploid species exhibiting a wide range of cell volume as a result of different endopolyploidy levels.


Plant material

The tested sample of 54 seed plant species comprised 24 annual and two biennial species, 17 perennial herbs and 11 wooden species (Table 1).

Table 1.  Endopolyploidization (expressed as ‘cycle value’), genome size and life time of 54 seed plant species
Cycle value
RootCoty- ledon
  1. The ‘cycle value’ indicates the mean number of endocycles per nucleus of a given organ. Genome size is taken from Barow & Meister (2002) or measured as described there. Life cycle is indicated as a, annual; b, biennial; p, perennial; pw, wooden perennial (tree or bush). Species in bold are endopolyploid.

GinkgoaceaeGinkgo biloba L. 21.58pw      0.00 0.00   
PinaceaeLarix decidua Mill. 25.73pw      0.00   0.00 
Abies concolor (Gord. et Glend.) Lind. ex Hildebrandt 36.12pw      0.00     
Picea abies (L) Karsten 38.63pw      0.00   0.00 
Pinus sylvestris L. 44.22pw      0.00     
RanunculaceaeAquilegia vulgaris L.  1.01p0.06
Anemone sylvestris L. 17.02p0.02 0.00
Anemone ranunculoides L. 36.83p0.04 0.02 0.07 
ChenopodiaceaeBeta vulgaris L.  1.84b      0.341.560.861.830.541.75
Atriplex rosea L.  2.12a      0.220.561.140.90 0.81
Spinacia oleracea L.  2.69a  2.300.53 1.580.380.941.361.421.451.45
UrticaceaeUrtica urens L.  1.07a      0.920.751.271.10 1.54
Parietaria officinalis L.  1.13p     0.600.580.790.710.86 0.84
Urtica dioica L.  2.34p
FagaceaeFagus sylvatica L.  1.30pw 0.000.24  0.050.03
Castanea sativa Mill.  1.96pw 0.07     
Quercus robur L.  2.18pw   0.120.06 0.030.13 0.06 0.00
RosaceaePhysocarpus opulifolius (L) Maxim.  0.71pw0. 0.03
Cydonia oblonga Mill.  1.98pw0. 0.00 0.00   
Duchesnea indica Focke  4.19p0.050.04  0.10 0.02 0.15   
FabaceaeTrifolium pratense L.  1.06p 1.01
Sophora japonica L.  1.34pw0.  0.11 0.05   
Phaseolus vulgaris L.  1.58a     0.600.250.310.870.87 1.36
Glycine max (L) Merr.  2.73a 1.05
Pisum sativum L.  9.07a0.750.441.010.240.640.470.140.170.430.540.651.78
Vicia faba L.26.21a0.860.240.740.180.440.550.210.080.690.59 1.77
BrassicaceaeArabidopsis thaliana (L) Heynh.  0.43a0.340.49  0.720.991.441.66
Sinapis arvensis L.  1.35a0.480.490.710.260.800.710.130.781.091.53 1.71
Raphanus sativus L.  1.38a,b1.050.981.780.320.980.950.140.950.741.330.681.43
Alliaria petiolata (M. Bieb.) Cavaraet Grande  2.70b0.180.320.760.21  0.300.460.760.79  
Brassica napus L.  2.95a,b0.430.680.670.260.850.790.130.930.481.39 1.27
CucurbitaceaeCucurbita moschata (Duch. ex Lam.)Duch. ex Poir.  0.97a   0.88  0.300.621.341.30 1.65
Cucumis sativus L.  1.03a1.091.261.55 1.43 0.421.35 1.98 1.74
Cucurbita pepo L.  1.18a1.121.371.51 1.47 0.580.291.701.25 1.24
Momordica charantia L.  1.43p0.630.631.21  1.060.881.231.151.591.231.67
SolanaceaeLycopersicon pimpinellifolium (Jusl.)Mill.  2.29a0.370.490.910.340.520.730.180.701.
Capsicum frutescens L.  7.37a 0.13
Nicotiana tabacum L.  9.77a0.290.220.530.270.49 0.110.23   0.72
LamiaceaeHyssopus officinalis L.  1.12p 0.01
Teucrium scorodonia L.  2.86p 0.01
Stachys grandiflora (Willd.) Benth. 12.47p0.
AsteraceaeHaplopappus gracilis (Nutt.) A.Gray  2.39a0.02  0.030.00
Lactuca sativa L.  6.61a,b0.02 0.010.05  0.030.02
Chrysanthemum multicolor Hyl. 32.55a0.00  0.010.00
AlliaceaeAllium ledebourianum Schult. etSchult. f.17.46p     0.480.49   0.85 
Allium cepa L.33.69p      0.260.36  0.480.62
Allium ursinum L.62.08p0.76 0.670.580.590.970.07 0.740.941.15 
Allium ampeloprasum L. s.l.65.48p      0.140.21   0.27
LiliaceaeFritillaria uva-vulpis Rix165.82p0.04 0.000.09  0.00     
PoaceaeOryza sativa ssp. japonica L.  1.18p      0.000.00   0.00
Zea mays L.  5.92a   0.53  0.200.71   0.87
Hordeum vulgare L.10.27a      0.260.18   0.18
Secale cereale L.16.01a   0.17
Triticum aestivum L.32.82a      0.050.14  0.080.13

The plants were grown in the greenhouse in pots containing standard potting soil. Samples from gymnosperms, Rosaceae, Fagaceae, and the species Sophora japonica L. (Fabaceae), Alliaria petiolata (M. Bieb.) Cavara et Grande (Brassicaceae), Allium ursinum L. (Alliaceae), and Anemone ranunculoides L. (Ranunulaceae) were harvested from plants growing outside.

As far as available, cotyledons, lower leaves (generally the first leaves of a plant), upper leaves and the corresponding leaf stalks were analysed for three to five individuals per species. Upper leaves and leaf stalks were generally taken from shoot parts directly below the inflorescence or at least from the topmost parts of the shoot. In addition, for at least one species of most families different parts of the flowers, upper parts of stems as well as roots were included. Probes from roots mainly consisted of differentiated tissues, but might have contained low amounts of meristematic cells since root tips had not been removed.

In total, more than 1700 samples were analysed for endopolyploidization.

Preparation, flow cytometric analysis and sorting of nuclear suspensions

Up to 100 mg fresh plant material was chopped in a pre-cooled Petri dish with a razor blade in about 1 mL ice-cold Galbraith buffer (45 mm magnesium chloride, 30 mm sodium citrate, 20 mm 4-morpholinepropane sulfonate and 0.1%[w/v] Triton X-100 according to Galbraith et al. 1983) modified by adding 5% (w/v) polyvinylpyrrolidone 25. The resulting suspension was filtered through a 35 µm nylon mesh and supplemented either with 50 µg mL−1 propidium iodide (Molecular Probes, Eugene, OR, USA) and 50 µg mL−1 DNase-free RNase (Boehringer Ingelheim Bioproducts Partnership, Heidelberg, Germany) or with 1 µg mL−1 4′-6-diamidino-2-phenylindole (DAPI; Molecular Probes) for DNA staining.

The DNA content per nucleus was measured with a FACStarPLUS. flow cytometer (Becton Dickinson, San José, CA, USA) equipped with two argon lasers INNOVA 90–5 (Coherent, Palo Alto, CA, USA) using the analysis programme CellQuest. Propidium iodide fluorescence was excited with 500 mW at 514 nm and measured in the FL1-channel using a 630 nm band-pass filter. The fluorescence of DAPI was excited with 200 mW in the UV range and measured in the FL1-channel using a 450 nm band-pass filter. The coefficient of variation (CV) of the histogram peaks depended strongly on the species and organ analysed, but were typically in the range from 3.5% (Cucumis sativus) to 7.0%(Quercus robur) on the base of a linear scale. In some gymnosperms, such as Larix decidua, it reached about 10.5%. Due to the preparation of nuclear suspensions by chopping entire organs, the data did not discriminate between particular tissues of the investigated organs.

For microphotometric investigation, 100 nuclei each were sorted in the Normal-R mode onto microscopic slides. Sorting was performed with a 100 µm nozzle as part of Becton Dickinson Macrosort accessory using the Sort Enhancement Module with a frequency of 15 000 s−1. The nuclei were selected by gating in the histogram and additionally in the fluorescence/side scatter dot plot in order to exclude unspecific fluorescence.

Analysis of flow cytometric histograms

For evaluation, the histograms were plotted with a logarithmic intensity scale (x-axis). This has the advantage that all peaks of a histogram are of the same width (Givan 2001). Therefore, the peak height is directly proportional to the number of nuclei of the corresponding ploidy level.

To reduce counts resulting from fluorescent debris, gates were set in the fluorescence intensity/side scatter density plot. In this plot, the region of nuclei is clearly determined by size and DNA content of nuclei whereas signals from debris are randomly distributed. Usually 10 000 nuclei were measured within a gate.

Some species showed a small peak in the histogram at the 4C position, just above the level of debris, that comprised on average about 3% of all the nuclei measured (Fig. 1a). To determine whether this peak was caused by 2C-nuclei sticking together, nuclei of this peak from leaves of four species (Chrysanthemum multicolor Hyl. [Asteraceae], Ginkgo biloba L. [Ginkgoaceae], Hyssopus officinalis L. [Lamiaceae] and Haplopappus gracilis[Nutt.] A.Gray [Asteraeceae]) were sorted and re-analysed with respect to their fluorescence intensity by a microscope photometer unit MPM 400 in connection with a microscope Axioskop (Carl Zeiss, Oberkochen, Germany). Sorted 2C-nuclei of the corresponding species were used as a standard. For only 50–70% of the sorted nuclei of the 4C peak the DNA content could be confirmed, the remaining nuclei revealed a DNA content of 2C. Therefore, only 1.5–2% of nuclei in the histogram actually represent 4C nuclei. Because of this low proportion of real 4C nuclei the corresponding species (all tested Lamiaceae and Asteraceae as well as two species of both Ranunculaceae and Rosaceae) are not considered to exhibit endopolyploidization.

Figure 1.

Logarithmic histograms of different cycle values. (a) Lactuca sativa, first leaf, cycle value = 0.018; (b) Vicia faba, first leaf, cycle value = 0.101; (c) Cucumis sativus, cotyledon, cycle value = 1.694.

In addition to 2C-nuclei sticking together, the 4C peak might contain a few G2 nuclei derived from meristematic cells of the bundle sheaths (Jacqmard et al. 1999).

Within the carpels of almost all species analysed, high 2C- and low 4C-peaks were found. By means of the programme ModFit (Verity Software House, Topsham, ME, USA) for cell cycle analysis we could prove the presence of S-phase nuclei, indicating that the 4C fraction represents mitotically active G2 nuclei. Later on, these nuclei may endoreduplicate since nuclei exceeding the 4C level were found in hulls of Vicia faba L and Pisum sativum L. (both Fabaceae).

Calculation of ploidy level

As a measure of endopolyploidization Engelen-Eigles, Jones & Phillips (2000) and Mishiba & Mii (2000) calculated the mean C-value (1C corresponds to the DNA content of reduced gametes) indicating the mean DNA content per nucleus. A disadvantage of this parameter is the overemphasis of the high ploidy levels because of the exponential character of the different ploidy steps.

For this reason, we defined the ‘cycle value’ indicating the mean number of endoreduplication cycles per nucleus.

The ‘cycle value’ is calculated from the number of nuclei of each represented ploidy level multiplied by the number of endoreduplication cycles necessary to reach the corresponding ploidy level. The sum of the resulting products is divided by the total number of nuclei measured.

Cycle value = (0 · n2C + 1 · n4C + 2 · n8C + 3 · n16C . . . )/ (n2C + n4C + n8C + n16C . . . ) where n2C,n4C, n8C . . . are the numbers of nuclei with the corresponding C-value (2C, 4C, 8C, . . . )

For illustration, logarithmic histograms for typical cycle values are shown in Fig. 1.

Statistical analysis

As the data are not normally distributed we calculated correlations by means of the Spearman test using the programme SigmaStat (Erkrath, Germany). A three-way analysis of variance with covariate (ancova) was applied using the programme SPSS 10.0 (SPSS Inc. 1999). According to Lindman (1974) this is legitimate because the F-test is robust against deviations from the normal distribution. As it was not feasible to measure all possible combinations family/organ, some of the cells in the data scheme are empty. Therefore, we analysed only the main effects (no interactions which are especially sensitive against missing combinations) and used type IV of sums of squares which is suitable for this situation (SPSS Inc. 1999).


Compilation of all data reveals a weak negative correlation between genome size and endopolyploidization

The ‘cycle values’ (see Materials and methods) of the tested organs of all species are listed in Table 1. An organ with a ‘cycle value’ below 0.1 was considered not to be endopolyploid. In most species showing endopolyploidization, it occurred in all organs tested except carpels. For this reason, these species are called endopolyploid species in the following. Only in Duchesnea indica Focke (Rosaceae) was endopolyploidy restricted to leaf and flower stalks where it occurred to a low degree. It is not clear whether the weak 4C peak in the first leaf of Fagus sylvatica L and Quercus robur L. (both Fagaceae) really contains endopolyploid nuclei. As the leaves were taken from plants at a very early stage of development, they possibly were not fully differentiated and might have contained meristematic cells. In Sophora japonica, only upper leaves narrowly exceeded the endopolyploidy cut-off value. Therefore, this species may also be considered not to be endopolyploid. In the endopolyploid species Glycine max (L) Merr. (Fabaceae), Capsicum frutescence L. (Solanaceae), Allium ursinum and Triticum aestivum L. (Poaceae) upper leaves revealed no endopolyploidy. In the latter, this is also true for roots. The lower leaves of the endopolyploid Vicia faba showed no endopolyploidization whereas in Aquilegia vulgaris L. (Ranunculaceae) only petals and stamina were without endopolyploidy.

In general, species with a small genome tend to be highly endopolyploid, for example, species of Brassicaceae, Cucurbitaceae and Chenopodiaceae. However, some species with a small genome showed none or only a low degree of endopolyploidization (e.g. Haplopappus gracilis, Hyssopus officinalis, Aquilegia vulgaris, Oryza sativa L. [Poaceae], the analysed Rosaceae and Fagaceae), whereas some species revealed endopolyploidy in spite of their large genomes such as Alliaceae and some species of Fabaceae and Poaceae (Table 1; Fig. 2). Taken together, a weak but significant negative correlation between genome size and endopolyploidization results from the total data set of our compilation (r = −0.253, P < 0.001).

Figure 2.

Endopolyploidization within families. Each dot indicates the endopolyploidization of one organ within one species. Each dot column represents different organs of the same species (Table 1). The ‘cycle value’ indicates the mean number of endoreduplication cycles per nucleus of an organ (see text). The dashed line indicates the endopolyploidization cut-off value (cycle value = 0.1). ‘r’ is Spearman's correlation coefficient (parentheses: correlation doubtful since no unequivocal endopolyploidy was measured, see Materials and methods). In gymnosperms no endopolyploidy was detectable. For Liliaceae, no figure is shown since only one species was measured that showed no endopolyploidy. Significance levels: *≤ 0.05, ** ≤ 0.01, *** ≤ 0.001.

Endopolyploidization is differently expressed in the different families

Figure 2 shows the relationship between ‘cycle value’ and genome size for each tested family. Except for Ranunculaceae, Rosaceae, Fabaceae and Poaceae, either all or none of the members of a family are endopolyploid, independent of the differences in the DNA content between the member species. Among Ranunculaceae and Rosaceae, only one species each (Aquilegia vulgaris, Duchesnea indica) showed endopolyploidization at a low level. Among the five tested species of Poaceae, only Oryza sativa was found to exhibit no endopolyploidization although it has the smallest genome of these five species. The tree-like Sophora japonica was the only species of the tested Fabaceae without endopolyploidy. These data suggest a strong correlation between taxonomic position and endopolyploidization of a species.

To exclude the superposition of the effect of genome size on a species’ endopolyploidization by its taxonomic placement we calculated the correlation between genome size and endopolyploidization separately for each family (Fig. 2). This calculation resulted in a significant negative correlation for five out of 14 angiosperm families. In one of these (Asteraceae) no and in another one (Ranunculaceae) only one species was unequivocally endopolyploid. In both families, the correlation is questionable because 2C nuclei sticking together or 4C nuclei of meristematic G2 cells might be responsible for the low 4C peaks (see Materials and methods). The observed significant correlation is therefore certain for only three out of the 14 tested angiosperm families (Brassicaceae, Urticaceae and Solanaceae). For Rosaceae, the Spearman test resulted in a moderate positive correlation, because the only endopolyploid species (Duchesnea indica) among the three tested species of this family is the one with the largest genome (see below).

All five tested gymnosperms were combined within one figure because no endopolyploidization was detectable. For the Liliaceae, only Fritillaria uva-vulpis Rix was tested which showed no endopolyploidization.

Endopolyploidization is differently expressed in species of different life cycle types

The graphic representation of endopolyploidization and genome size of the tested species separated as to short-lived herbs, perennial herbs and wooden species, respectively (Fig. 3), reveals differences between these life-cycle types. Annuals, facultative biennials, overwintering annuals and obligate biennials (see Table 1) were analysed together as no obvious difference regarding endopolyploidization was discernible. 23 out of 26 annual and biennial species are endopolyploid, whereas this is the case for only 10 of the 17 perennial herbs. All three non-endopolyploid exceptions among the annual species belong to the Asteraceae. Two of the tested perennial herbs (Duchesnea indica, Aquilegia vulgaris) show only a very low degree of endopolyploidization. No endopolyploidy was found in the tested wooden species (Fagaceae, Gymnospermae, the two Rosaceae Cydonia oblonga Mill. and Physocarpus opulifolius[L.] Maxim., and the Fabaceae species Sophora japonica).

Figure 3.

Endopolyploidization in species of different life cycles. Short-lived species comprise annuals and biennials. Each dot indicates the endopolyploidization of one organ within one species. Each dot column represents different organs of the same species (Table 1). The ‘cycle value’ indicates the mean number of endoreduplication cycles per nucleus of an organ (see text). The dashed line indicates the endopolyploidization cut-off value (cycle value = 0.1). ‘r’ is Spearman's correlation coefficient. Significance levels: *≤ 0.05, ** ≤ 0.01, *** ≤ 0.001

The statistical analysis resulted in a significant negative correlation between genome size and endopolyploidization for the short-lived species (r = −0.586, P≤ 0.001), but for the other two life-cycle types the coefficient of correlation was nearly zero (perennial herbs: r = −0.086, P ≤ 0.05; wooden species: r = 0.018, P > 0.05).

Endopolyploidization is differently expressed for the different organs

On average, the highest degree of endopolyploidization was found for cotyledons and leaf stalks (Fig. 4). Lower leaves and leaf stalks showed a higher degree of endopolyploidization than the upper ones. Carpels are probably not endopolyploid (see Materials and methods).

Figure 4.

Endopolyploidization within organs. Each dot indicates the endopolyploidization of the organ in one of the species (Table 1). The ‘cycle value’ indicates the mean number of endoreduplication cycles per nucleus of an organ (see text). The dashed line indicates the endopolyploidization cut-off value (cycle value = 0.1). ‘r’ is Spearman's correlation coefficient. Significance levels: *≤ 0.05, ** ≤ 0.01, *** ≤ 0.001

For all organs except carpels, a weak to moderate negative correlation was discernible which was not significant in stamina and flower stalks. The significant coefficients of correlation ranged from r = −0.454 to −0.169. The strongest correlation was found for leaves, leaf stalks and sepals.

Family affiliation, life cycle, organ type and genome size are related to endopolyploidization to a different extent

To quantify the correlation of the factors ‘family’, ‘life cycle’ and ‘organ type’ as well as genome size with endopolyploidization, we applied a three-way analysis of variance with covariate (ancova). This analysis confirmed the main impact of family affiliation on occurrence and degree of endopolyploidization as well as the lesser importance of genome size in this regard (Table 2). Organ type and life cycle have a clearly lower impact on endopolyploidization than the factor family affiliation, but a noticeable higher one than the genome size.

Table 2.  Three-way analysis of variance with covariate for the impact of the factors ‘family affiliation’, ‘organ’, ‘genome size’ and ‘life cycle’ on endopolyploidization
ParameterDegrees of freedomF
  1. Significance level: *≤ 0.05, **≤ 0.01, ***≤ 0.001.

Family affiliation15115.7***
Life cycle 344.1***
Genome size (covariance) 15.4*

For evaluation of these results, one has to bear in mind the relations between the factors considered within the experiment:

  • 1Genome size is correlated with the life cycle in the way, that plant species with a short life span (annuals and biennials) have on the average a small genome (Bennett 1972; Vinogradov 2001). Furthermore, the families differ in their average genome size. We tried to overcome this problem by choosing species with genome sizes as different as possible within each life cycle and within each family. Nevertheless, the significance of the factor ‘genome size’ might have been diminished since the statistical test assigned the life cycle- and family-specific portion of genome size to the factors ‘life cycle’ and ‘family’ without considering the relations between the genome size and the other two factors.
  • 2Another relation exists between family and life cycle: the majority of species within each family has the same life cycle (Table 1). Therefore, one could suspect, that the apparent influence of the factor ‘family’ is actually caused by the factor ‘life cycle’. If this is the case, the factor ‘family’ should be dispensable, and removing it from the statistical analysis should not change the result essentially. However, the calculation of analysis of variance/covariance without the factor ‘family’ does not support this view: the overall F-value for the model halves and accordingly the residual variance doubles approximately.

This result shows clearly that the family affiliation is an essential factor for the extent of endopolyploidization beside the life cycle type. Some exceptions from the relation family versus life cycle type support this result. All tested Asteraceae are annuals but without endopolyploidy whereas the tested Alliaceae are perennials and endopolyploid. Likewise, two of the three endopolyploid Urticaceae are perennials (Parietaria officinalis, Urtica dioica) as well as one species of the three tested endopolyploid Cucurbitaceae (Momordica charantia) and one of the predominantly endopolyploid Fabaceae (Trifolium pratense).


The negative correlation between endopolyploidization and genome size claimed by Nagl (1976) cannot be considered as a universal feature of seed plants. Our data based on different organs of 54 plant species reveal that endopolyploidization is primarily related to their taxonomic position (Table 2; Fig. 2). In some taxonomic groups many species contain highly endopolyploidized tissues whereas this is not the case in other taxa.

A clear negative correlation between endopolyploidization and genome size could be found only for three of the 16 investigated families.

The degree of endopolyploidization differs between the different organs of a species.

For all organs except carpels a weak to moderate negative correlation between genome size and endopolyploidization results from our data if they are analysed separately for the different organ types.

Additionally, endopolyploidization is related to life cycle. Endopolyploidy is very frequently found in annual and biennial species and also in some perennial herbs whereas it seems to be absent in wooden species.

Occurrence of endopolyploidy in seed plants is determined genetically

The tight relationship between taxonomic position and endopolyploidization we found, and as already proposed by Tschermak-Woess (1956), D’Amato (1964), Nagl (1976) and Olszewska & Osiecka (1982), suggests a genetic determination of a species’ capacity to endopolyploidize. Nevertheless, families belonging to closely related clades may differ completely with respect to this feature, for instance, Solanaceae and Lamiaceae, Fagaceae and Cucurbitaceae and Rosaceae and Fabaceae.

Within the analysed Fabaceae, Sophora japonica (the only one without endopolyploidy among the tested Fabaceae) belongs to the tribe Sophoreae consisting of mainly wooden tropical and subtropical species whereas the other species belong to the tribes Vicieae, Trifolieae and Phaseoleae of which the members are predominantly herbs of temperate climates. Likewise, within the tested Poaceae, the three species with similar endopolyploidy pattern Triticum aestivum, Secale cereale and Hordeum vulgare belong to the Pooideae, whereas Zea mays and Oryza sativa with higher and without endopolyploidization, respectively, are assigned to two different subfamilies.

Further results on endopolyploidization in species, which we analysed for different purposes but not in detail with regard to endopolyploidy, confirm our results that the degree of endopolyploidization is family specific or at least very similar for closely related species. The upper leaves of five species of Lamiaceae (Glecoma hederacea L., Lamium purpureum L., Ocimum basilicum L., Galeobdolon luteum Huds. and Betonica officinalis L.), one species of Asteraceae (Bellis perennis L.), two Ranunculaceae (Ranunculus ficaria L. and Caltha palustris L.) and two Rosaceae (Potentilla reptans L. and Fragaria viridis[Duchesne] Weston) showed no endopolyploidization as expected for these species according to our results in Table 1. Leaves of 10 species of Brassicaceae (five Arabidopsis species besides A. thaliana, Teesdalia nudicaulis[L.] R. Br., T. coronopifolia[J.P. Bergeret] Thell., Cardamine amara L., Capsella rubella Reuter, Erysimum pulchellum J. Gray) and Allium scorodoprasum L. (Alliaceae) were endopolyploid as one could expect pursuant to our results.

Two species show that there are also exceptions to a family's predominant endopolyploidization behaviour: Erysimum helveticum Jacq. (Brassicaceae) did not show endopolyploidy in upper leaves whereas Fragaria vesca L. (Rosaceae) showed endopolyploidy in leaves to a low degree.

The degree of endopolyploidization differs between the organs of endopolyploid species

The differences between different organ types as to their degree of endopolyploidization, described by several authors (De Rocher et al. 1990; Galbraith et al. 1991; Gilissen et al. 1993; Smulders et al. 1994), are confirmed by our data. Moreover, the different degrees of endopolyploidization of mature organs of the same type (lower leaves and leaf stalks reach a higher endopolyploidization than upper leaves and leaf stalks) in Arabidopsis thaliana (Galbraith et al. 1991) are apparently a common phenomenon in seed plants (Fig. 4).

These differences do not reflect the different ages of organs but rather their ontogenetic appearance (lower versus upper) and are apparently genetically fixed. The level of endopolyploidization does not increase within an organ once it is fully developed (De Rocher et al. 1990; Gilissen et al. 1993). Assuming a positive correlation between endopolyploidy level and cell volume in plants (Bradley 1954; Melaragno et al. 1993; see below) a further endopolyploidization in fully developed organs is unlikely since it would lead to a further cell expansion.

Although Galbraith et al. (1991) did not find endopolyploidy in flower buds of Arabidopsis thaliana we found high degrees of endopolyploidization in fully differentiated flower organs, except carpels, for A. thaliana and other endopolyploid species. Hence, endopolyploidy is not excluded from floral structures in general to avoid the potential production of polyploid gametophytes as assumed by Galbraith et al. (1991).

The negative correlation of genome size and endopolyploidization may be caused by the correlation between genome size and life cycle and the growth features of large cells

A compensation of a low amount of DNA by endopolyploidization in species with small genomes to enable certain cell functions is unlikely because of the proportionality of DNA content and cell volume of endopolyploid cells that means an equal ratio of gene products and cell volume. A positive correlation between DNA content and cell volume was demonstrated for endopolyploid pith cells of Nicotiana tomentosa with ploidy levels ranging from 4C to 16C (Bradley 1954) and for epidermal pavement cells of Arabidopsis thaliana leaves with a ploidy range from 2C to 16C (Melaragno et al. 1993). Furthermore, the size of trichomes of Arabidopsis thaliana and the number of their branches depend on their ploidy level (Hülskamp, Miséra & Jürgens 1994). Moreover, in Medicago truncatula the reduced amount of transcripts of the gene ccs52 involved in the switch from mitotic cycles to endocycles leads to a decreased level of endopolyploidy and cell size in roots, cotyledons, petioles and hypocotyls (Kondorosi et al. 2000).

However, comparing mean size and mean ploidy levels of mature root cortex cells of 18 ecotypes of Arabidopsis thaliana, Beemster et al. (2002) showed that even among ecotypes of one species the relationship between these two may vary. In the KRP2-mutant of Arabidopsis thaliana, in which cell division is inhibited and endoreduplication is suppressed in older leaves, cells are enlarged in comparison with the wild type in the absence of enhanced endoreduplication (De Veylder et al. 2001). Likewise, overexpression of the Arabidopsis thaliana cell cycle regulator Cdc2a in tobacco led to fewer but larger cells in leaves although the mean endopolyploidy was the same as in wild-type tobacco (Hemerly et al. 1995). Possibly, the point is that there is a close relation between DNA-content and volume of cytoplasm rather than cell volume. The latter may vary to a considerable extent depending on the cell type, organ, ecotype or mutant caused by the vacuole, which contributes to a high degree to the cell volume. Already Tschermak-Woess (1956) emphasized this problem. However, for meristematic cells, which do not yet have a vacuole, Price, Sparrow & Nauman (1973) reported a close relation between genome size and cell volume for 14 herbaceous angiosperms. Furthermore, within each single mutant and accession investigated by Beemster et al. (2002), De Veylder et al. (2001) and Hemerly et al. (1995) the DNA content/cell volume-relation may hold true.

Therefore, as an alternative to a compensation of a low amount of DNA by endopolyploidization (Nagl 1976), considering the correlation between life cycle and endopolyploidization and life cycle and genome size on the one hand and the growth feature of species with large genomes on the other (see below; Grime & Mowforth 1982 and Grime et al. 1985), one might speculate that endopolyploid species combine the advantages of small genomes and large cells to speed up development.

  • 1Advantages of a small genome. Annual and biennial plant species preferentially possess small genomes (Bennett 1972; Vinogradov 2001) since species with small genomes tend to have short generation times. This seems reasonable on the basis of a positive correlation between genome size on the one hand and somatic cell cycle time, duration of meiosis and pollen maturation on the other (Bennett 1972). Because of a shorter cell cycle time, plant species with a small genome are expected to grow faster under supporting conditions (e.g. high temperatures) than species with a large genome when growing mainly by cell division (quasi exponentially) and subsequent cell expansion.
  • 2Advantages of large genomes. Plant species growing at higher latitude tend to have larger genomes (Levin & Funderburg 1979; Bennett 1976). Grime & Mowforth (1982) and Grime et al. (1985) could show that this is due to the differential effect of low temperatures on cell division and cell expansion. Although at low temperatures mitosis is often inhibited, this is not the case for cell expansion. Accordingly, investigating ploidy levels and DNA amounts of 12 Tulipa, 13 Hyacinthus, and 16 Narcissus cultivars, Brandham & West (1993) demonstrated that a similar genome size, in this case a large amount of DNA, due to either a large diploid genome size or acquired by polyploidization, is optimal for evolutionary adaptation of genera, which are not closely related, to environmental conditions that inhibit mitosis but not cell expansion since fewer cells for the same growth rate are needed.

We speculate that endopolyploid species are able to develop faster than species without endopolyploidy with the same genome size by combining these advantages. In endopolyploid species the corresponding organs may be able to grow faster, since initial growth by cell division is followed by extensive cell expansion, mediated by endopolyploidization. At low temperatures, endopolyploid species with small genomes should grow as fast as non-endopolyploid species with large genomes since they may reach the same cell size by expansion. At temperatures that are optimal for mitosis, the former should grow faster by combining short cell cycles and endopolyploidy-mediated cell expansion. This hypothesis may be tested by systematic investigation of growth of endopolyploid versus non-endopolyploid species of different genome sizes under different conditions.

Fast growth and development of endopolyploid species might get further support since less cell surface material due to larger cells (Barlow 1978), no spindle formation, no chromosome condensation and decondensation, no breakdown and reconstruction of the nuclear envelope is needed and the transcription is maintained during endoreduplication (Nagl 1978). In addition, generation time is shortened by faster meiosis and pollen maturation in species with small genomes.

This conclusion corresponds to the suggestion of Cavalier-Smith (1978) that the most fundamental way organisms adapt to varying r- and K-strategy (which are, according to Grime 1974, 1977 and 1988, analogous to the C-, S- and R-strategies in the plant kingdom) is by evolving particular cell volumes and cell growth rates.

The importance of endopolyploidization for higher growth rate is further indicated by hypocotyls grown in the dark and in light. Hypocotyls of Lupinus albus grown in the dark are four times longer than those grown in light whereas cell number is the same in both (Giles & Myers 1964). This is apparently due to endopolyploidization since DNA content per hypocotyl is increased. In addition, for Pisum sativum (Van Oostveldt & Van Parijs 1975) and Arabidopsis thaliana (Gendreau et al. 1998) a higher endopolyploidization level was found in etiolated than in non-etiolated epicotyls and hypocotyls, respectively. Hence, the potential to grow by endopolyploidization seems to be advantageous for etiolated seedlings since it leads to elongated shoots by fast growth. This might be crucial under natural conditions in order to pass the soil surface for starting photosynthesis.

An advantage of endopolyploidy for fast growth could even explain the different degree of endopolyploidization between organs of the lower and upper parts of a plant since the lower parts often develop during earlier seasons and thus at lower temperatures in temperate climes.

In some species, we found endopolyploidy only to a low degree and not in all organs where it normally occurs. Therefore, in these species endopolyploidization barely serves for a fast development. Identification of individual cells that are actually endopolyploid could provide clues with respect to the function of endopolyploidy within these species, which could be generation of certain morphological structures (Pyke, Marrison & Leech 1991; Kondorosi et al. 2000) or support of high metabolic activity as in suspensor cells (Nagl 1978).

Considering the significance of endopolyploidization for growth and development of plants, the weak correlation of endopolyploidization with genome size and life cycle is astonishing. This might be explained by the superposition of these two correlations by the life strategy of the corresponding species (the strategy a species embarks on to compete successfully for a ecological niche), which includes more ecological features than just the life cycle.

Life strategy may possibly superpose the correlation of endopolyploidization with life cycle and genome size

Twenty-three out of the 26 tested short-lived species revealed endopolyploidy, and all tested bushes and trees did not. This agrees well with the assumed relationship between endopolyploidization and life cycle. Nevertheless, more than one-half of the tested perennial herbs are endopolyploid. Therefore, endopolyploidy is possibly not only important for species with short generation times, which have to complete the whole life cycle from germination to seed production within one season. A fast development mediated by endopolyploidy might also be advantageous for perennial herbs that can exploit convenient growing conditions available only temporarily. In these cases, it is rather life strategy in general than life cycle as a special part of life strategy that drives exhibition of endopolyploidization. Four of the five tested Central European endopolyploid perennial herbs mainly grow at frequently disturbed locations (Trifolium pratense: anthropo-zoogenic heathlands and grasslands; Aquilegia vulgaris, Parietaria officinalis, Urtica dioica: herbaceous vegetation of often disturbed locations; Ellenberg et al. 1992). They have to complete growing, flowering and seed production within a limited period or have to be able to regenerate quickly after having been disturbed. The fifth tested Central European endopolyploid perennial herb, Allium ursinum, exhibiting a relatively high endopolyploidization in addition to its large genome, has to develop quickly before deciduous trees shadow its habitat during late spring and summer. Anemone ranunculoides, which is found in a similar habitat to Allium ursinum and which has a relatively large genome, too, but no endopolyploidy, starts growing and flowering much earlier in the year and stays green until approximately the same time as Allium ursinum.

Why do not all species growing in often disturbed habitats exhibit endopolyploidy since it appears to be so advantageous (e.g. the tested Asteraceae and Lamiaceae)? A second factor, namely a species’ nutritive demand, may favour endopolyploidization of species of the same life cycle type to different degrees. Ellenberg et al. (1992) divided the seed plant species of Central Europe in respect of this feature into nine classes. This classification reveals that most of the tested endopolyploid Central European species have higher nutritive demands than most species without endopolyploidy. Eight endopolyploid herbaceous perennial species are at least considered to be more abundant in nitrogen-rich habitats than in mediocre habitats. Trifolium pratense (Fabaceae) is not characterized by Ellenberg et al. (1992) in respect to its nutritive demands but an adequate nitrogen supply can be assumed as in general for this family because of nitrogen-fixing endosymbionts in root nodules. Arabidopsis thaliana is the only tested endopolyploid Central European species with low nutritive demands. On the other hand, four of the tested Central European species without endopolyploidy are more abundant in nitrogen-poor habitats (if one accepts Lactuca serriola as ancestor of L. sativa and therewith the classification of L. serriola). This is also true for Aquilegia vulgaris, which exhibits endopolyploidization only to a low extent. Anemone ranunculoides is the only tested herbaceous perennial Central European species without endopolyploidization but high nutritive demands.

When endopolyploid species indeed may develop faster than species without endopolyploidization (also Melaragno et al. 1993 discussed a more rapid growth by endopolyploidization) it seems to be understandable that they preferentially occur in habitats that require a fast completion of growth, flowering and seed production and support fast development by optimal nutrient supply. Apparently, in such habitats, the ability of endopolyploid species to make use of supporting conditions by fast development enables them to compete successfully with species without endopolyploidy. Investigation of genome size, endopolyploidization and phenology of species that are more or less restricted to a given habitat could reveal whether endopolyploidization is of ecological significance under defined growth conditions.

Similar life strategies of species of one family might cause the strong correlation between endopolyploidy and family affiliation

Within many plant families species tend to grow in similar habitats (Hodgeson 1986; Ellenberg et al. 1992). For instance, in Central Europe, arable farmland is the habitat richest in species of Brassicaceae, Chenopodiaceae and Solanaceae (Ellenberg et al. 1992), families, which revealed in our experiments very high degrees of endopolyploidization. On the other hand, wood is the habitat richest in species of Ranunculaceae, Rosaceae and rich in Lamiaceae, families, which show low or no endopolyploidization. Similar ecological preferences of species of the same family imply that it is rather the predominant life strategy of a family than the family affiliation itself that causes the strong correlation between the factor ‘family’ and endopolyploidization. If this would be the case, species that differ from the predominant life strategy of a family should also differ from a family's main endopolyploidization behaviour. In fact, this is true for three of the tested species, namely, Aquilegia vulgaris (Ranunculaceae), Duchesnea indica (Rosaceae) and Sophora japonica (Fabaceae). If one considers the relationship between life strategy and endopolyploidization and the apparent absence of a close relation between phylogeny and endopolyploidization it seems likely that the former favours endopolyploidy within closely related species.


We thank Ingo Schubert for helpful discussions, Konrad Bachmann, Fritz Matzk, Klaus Pistrick and Rigomar Rieger for critical reading of the manuscript, Katrin Menzel and her staff for qualified cultivation of the plants, and the team of the IPK germplasm collection for providing seed material. The work was supported by the Deutsche Forschungsgemeinschaft (ME 1083/2) and the Fond der Chemischen Industrie (to A.M).

Received 6 September 2002; received in revised form 3 October 2002; accepted for publication 10 October 2002