Partial Diploidization of Meiosis in Autotetraploid Arabidopsis thaliana
J. L. Santos, D. Alfaro, E. Sanchez-Moran, S. J. Armstrong, F. C. H. Franklin, G. H. Jones


Meiosis was analyzed cytogenetically in autotetraploids of Arabidopsis, including both established lines and newly generated autotetraploid plants. Fluorescent in situ hybridization with 5S and 45S rDNA probes was used to identify the different chromosomes at metaphase I of meiosis. Multivalents were observed frequently in all the lines analyzed, but there were significant differences in multivalent frequency not only between the newly generated tetraploids and the established lines but also among the different established lines. The new tetraploids showed high multivalent frequencies, exceeding the theoretical 66.66% predicted by the simple random-end pairing model, in some cases significantly, thus indicating that Arabidopsis autotetraploids have more than two autonomous pairing sites per chromosome, despite their small sizes. The established lines showed fewer multivalents than the new autotetraploids did, but the extent of this reduction was strongly line and chromosome dependent. One line in particular showed a large reduction in multivalents and a concomitant increase in bivalents, while the other lines showed lesser reductions in multivalents. The reduction in multivalents was not uniformly distributed across chromosomes. The smaller chromosomes, especially chromosomes 2 and 4, showed the most marked reductions while the largest chromosome (1) showed virtually no reduction compared to the new tetraploids. It is concluded that the established autotetraploid lines have undergone a partial diploidization of meiosis, but not necessarily genetical diploidization, since their creation. Possible mechanisms for the resulting change in meiotic chromosome behavior are discussed.

THE analysis of meiotic chromosome behavior in polyploids has considerable practical value in terms of understanding, and possibly moderating, the undesirable effects of polyploidy on fertility and stability (Evans 1981; Gillies 1989). Furthermore, it has long been recognized that the special situation found in autopolyploids can give unique information of a fundamental nature relating to the organization of chromosome pairing and synapsis during meiosis (Sybenga 1975). When more than two homologous chromosomes are present, as in autopolyploids or polysomics, it is usually found that they are restricted to pairwise association, but different pairwise combinations of chromosomes may associate at different points along chromosomes. When this happens, multivalents containing pairing partner switches (PPSs) arise in prophase I, and these multivalents may persist to metaphase I, provided that they have the requisite numbers and distributions of chiasmata to bind them together. The analysis of synaptic configurations in prophase I by electron microscopical analysis of synaptonemal complexes can give direct and detailed information on the numbers and distributions of PPSs and hence on the numbers of autonomous pairing sites (APSs; Callow and Gladwell 1984; Loidl and Jones 1986; Jones and Vincent 1994; Santoset al. 1995). However, even without prophase I observations, it is possible to make valuable inferences on meiotic pairing/synapsis from observations on metaphase I chromosome configurations (Diezet al. 2001).

The occurrence of autopolyploid variants of Arabidopsis has been reviewed recently by Koornneef et al. (2003). Autotetraploid plants of Arabidopsis have been recovered following colchicine treatment (Bouharmont and Van den Hende 1968; Bouharmont 1969) and regeneration from tissue culture (Morris and Altmann 1994) as well as spontaneous occurrence in certain accessions (Heslop-Harrison and Maluszynska 1994). They have been employed in numerous studies of the developmental, physiological, and morphological effects of chromosome number change (Bronckers 1963; Bouharmont and Macé 1972; Karczet al. 2000). Surprisingly, there have been very few attempts to analyze their meiotic chromosome behavior (Weiss and Maluszynska 2000) although the potential advantages of doing so in this model organism are considerable. Arabidopsis is an important plant model for meiosis research and, as a result, the knowledge base and resources for investigating meiotic phenomena and effects, including the isolation and characterization of several meiotic genes, are rapidly increasing. In addition, cytogenetical techniques for the analysis of meiosis in Arabidopsis, including the application of fluorescence in situ hybridization (FISH) for chromosome identification, have improved dramatically in recent years (Franszet al. 1998; Sanchez-Moranet al. 2001).

In the present study, these cytogenetic methodologies have been applied to the analysis of chromosome configurations at metaphase I in autotetraploid Arabidopsis. The study includes a comparison of four established autotetraploid lines (E lines) that have been maintained for a minimum of 13 generations and newly produced colchicine-generated autotetraploid plants (C). The analysis shows conclusively that multivalent formation is of frequent occurrence in all the autotetraploid Arabidopsis material analyzed. However, there is clear evidence of increased bivalent frequency in the established lines, indicating that a partial diploidization of meiosis has occurred in these lines, although the extent of this differs among lines and is also markedly chromosome dependent.


Plant material: Established (E) autotetraploid lines of Arabidopsis were obtained from the Arabidopsis Biological Resource Center (ABRC) at Ohio State University. They correspond to stock numbers CS 3151 (E1), CS 3427 (E2), CS 3245 (E3), and CS 3432 (E4). All four E lines originated in the Col accession and were donated by G. Rédei. To verify that these lines all share the same accession background (Col), microsatellite analysis was conducted on the four E lines and on the parental (Col) line that was used to generate the new (C) autotetraploid plants. Five microsatellite loci were analyzed, one from each of the five Arabidopsis chromosomes (chromosome 1, F23A5; chromosome 2, T9J23; chromosome 3, TH620B; chromosome 4, nga8; chromosome 5, nga139; All five microsatellite loci were found to be monomorphic across the five lines analyzed, thus confirming that they are extremely likely to share the same accession background. The microsatellite gel images may be viewed at http:/

One of the E lines (CS 3432) is recorded in the ABRC stock listing as having arisen by spontaneous chromosome doubling. The stock listing does not record whether the other three E lines arose spontaneously or were induced by colchicine treatment; however, this should have no bearing on their subsequent meiotic behavior. These lines were subsequently maintained by self-pollination for at least 13 generations (G. P. Rédei, personal communication). To generate new autotetraploids, young plants of the Columbia accession were colchicine treated at the preflowering rosette stage by placing 1 drop (7 μl) of 0.25% colchicine solution on the rosettes. The flowering stems of surviving plants were wholly tetraploid. These new autotetraploids are referred to, for convenience, as the C line, but it is important to realize that this is not strictly equivalent to the E lines. Each C plant is the result of an independent chromosome doubling event, whereas the E lines are most probably each derived from a single initial polyploidization episode.

Seeds of all five lines were sown directly onto soil-less compost and grown to flowering in a constant environment chamber at a temperature of 18° and a day length of 16 hr.

Fixation: Immature flower buds were detached from the plants and fixed in Carnoy’s fixative (6 ethanol:3 chloroform:1 acetic acid). Fixed flower buds were stored in fixative at -20° until required.

Slide preparation: Air-dried spreads of pollen mother cells (PMC) were prepared according to the method of Fransz et al. (1998) and Armstrong et al. (1998), with minor modifications as described by Sanchez Moran et al. (2001). Fixed buds were washed in fixative (3 ethanol:1 glacial acetic acid), followed by citrate buffer (pH 4.5), and then incubated in enzyme mixture, 0.3% (w/v) pectolyase, 0.3% (w/v) cytohelicase, 0.3% (w/v) cellulase (all Sigma) in citrate buffer, for 1.5 hr at 37°. Replacing the mixture with ice-cold buffer stopped the reaction. Single buds were transferred to clean slides, together with a small volume of buffer, and macerated with a needle. A total of 10 μl of 60% acetic acid was added to the slide before placing on a hot plate at 45° for 1 min while stirring with a needle. A further 10 μl of 60% acetic acid was added to the slide off the hotplate before adding 200 μl of cold 3:1 fixative. The fixative was drained away, and the slide dried with a hair drier.

FISH: The FISH technique used was previously described by Armstrong et al. (1998) and Fransz et al. (1998). Slides were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; 4 μg/ml) in Vectashield antifade mounting medium (Vector, Burlingame, CA).

The following DNA probes were used:

  • Clone pTa71 (Gerlach and Bedbrook 1979) containing a 9-kb EcoRI fragment of Triticum aestivum consisting of the 18S-5.8S-25S rRNA genes and the spacer regions. This probe was directly labeled with Spectrum Green (Amersham, Buckinghamshire, UK) by nick translation following the manufacturer’s instructions (Boehringer Mannheim, Indianapolis).

  • Plasmid pCT4.2 containing the 5S rDNA gene from A. thaliana as a 500-bp insert cloned in pBlu. This probe was generated by PCR using M13 primers (Pharmacia). Biotin dUTP was incorporated in a secondary PCR reaction.

The FISH preparations were viewed with an epifluorescence microscope (Nikon E600) with filters for DAPI, TRITC, and FITC and equipped with a Quips image capture and analysis system (Applied Imaging International, Sunderland, Tyne and Wear, UK).

Statistical analysis: All statistical analyses were carried out using Minitab software. The experimental design permitted the use of a theoretical error (820.7/n with ∞ d.f.) in the two analyses of variance (Fisher and Yates 1963).


Chromosome identification: It has been shown previously (Franszet al. 1998; Sanchez-Moran et al. 2001, 2002) that 45S and 5S rDNA FISH probes, combined with chromosome morphology, uniquely identify each of the Arabidopsis chromosomes and chromosome arms (Figure 1). The short acrocentric chromosomes 2 and 4 both carry 45S sequences distally on their short arms, in all accessions that have been examined, coinciding with the locations of the nucleolus organizing regions (NORs). In addition, all accessions examined also have a 5S rDNA site located proximally on the short arm of chromosome 4, which serves to distinguish the two acrocentric chromosomes. A further invariant large 5S site occurs proximally on the shorter arm of the submetacentric chromosome 5, which serves to distinguish this chromosome from chromosome 1. Chromosome 3 is the smallest of the submetacentric/metacentric group of chromosomes (1, 3, and 5) and in the Col accession is distinguished by possession of a third 5S rDNA site located proximally in the short arm; this 5S site is much smaller than the similarly positioned site on chromosome 5.

Figure 1.

—Three representative examples (a-c) of FISH images of metaphase I cells from autotetraploid Arabidopsis to show quadrivalent (IV) and bivalent (II) configurations; superscript Arabic numerals identify the different chromosomes (1-5). Green signals identify 45S rDNA loci and red signals identify 5S rDNA loci. The accompanying line drawings (d-f) show how the FISH images have been interpreted.

Metaphase I configuration frequencies: Metaphase I configuration frequencies were recorded for each chromosome from 50 FISH-probed PMCs per line (Table 1). The chromosomes were predominantly associated as either bivalents or quadrivalents (Figure 1). Further examples of metaphase I configurations may be viewed at Occasionally a set of four chromosomes formed a trivalent + univalent, but this occurred relatively infrequently (19/1250 = 1.52%). In practice, for the purposes of this study, no distinction was made between quadrivalents and trivalents and they were simply grouped together as multivalents.

Because of limitations on the numbers of cells that could be scored from individual plants, data were collected from two plants per line. A preliminary analysis was conducted on part of the data set, restricted to give equal numbers of cells per plant per line (omitting line E1 because of severe inequality of numbers). Multivalent percentages were transformed to angles, and a semihierarchical (nested) analysis of variance was carried out (Table 2). This analysis showed that there were no significant differences in multivalent frequencies between plants within lines, and on the basis of this finding the individual plant data were combined to give estimates of multivalent frequencies based on 50 cells per line over the entire data set (Table 1).

It can be seen from Table 1 that multivalent frequencies per line, averaged over chromosomes, vary between 52.8 and 71.2% among the E lines, while C plants have, on average, 79.0% multivalents. An analysis of variance of this data set, again following angular transformation, confirms that the five lines included in this study differ significantly for multivalent frequency (Table 3). Furthermore, C plants have significantly higher multivalent frequencies than the E lines, while the E lines differ significantly from each other. This analysis also shows that, overall, multivalent frequency differs significantly among chromosomes. The “chromosomes × lines” interaction item is also significant, indicating that chromosomes 1-5 do not behave consistently over lines as regards multivalent formation. However, when this item is partitioned into its component parts, it emerges that the inconsistency is attributable to C vs. E lines; the chromosomes behave consistently over the four E lines.

These patterns of variation in multivalent frequencies are seen more clearly when the data are expressed graphically. In Figure 2, multivalent frequencies per chromosome are plotted against pachytene chromosome length (data of Franszet al. 1998) for each of the lines, revealing several points of interest and complementing the statistical analysis.

View this table:

Multivalents (M) and bivalent pairs (IIs) observed for each of the five chromosomes in 50 pollen mother cells per line

C plants show a consistently high multivalent frequency with relatively little variation between chromosomes (74-82%). In contrast, the four E lines show fewer multivalents and more bivalents, but the extent of this is highly chromosome and line dependent. It can be seen that the reduction in multivalent frequency in E lines compared to C preferentially affects the smaller chromosomes of the genome, particularly the short acrocentric chromosomes 2 and 4. The medium-sized submetacentric chromosomes 3 and 5 exhibit a less pronounced reduction in multivalents, restricted to only one or two E lines. On the other hand, the longest chromosome (1) shows no discernible reduction and all five lines have closely similar multivalent frequencies for this chromosome (76-86%).

In line with the ANOVA results, there are also clearly discernible differences among the E lines. Line E4 has the lowest overall multivalent frequency, affecting all chromosomes, except for chromosome 1, to a greater or lesser extent. Line E3 has the next lowest multivalent frequency, affecting chromosomes 2 and 4 particularly, but differing from line E4 in that chromosome 3 has a relatively high multivalent frequency, comparable to that seen in the new C plants.

Quadrivalent:bivalent ratios: The simplest model of chromosome pairing in autotetraploids, the so-called random-end pairing model, assumes that pairing initiation is restricted to chromosome ends and that there is random choice of pairing partners among the four homologous chromosomes at the two ends. Given these assumptions, the model predicts that multivalent formation at prophase I is twice as likely as bivalent formation and therefore the ratio of multivalents to bivalent pairs will be 2:1 (equivalent to a 1:1 ratio of multivalents to bivalents); this can also be expressed as a multivalent frequency of 66.66% (Sybenga 1975; Jackson and Casey 1982; Jones and Vincent 1994; Jones 1994). The observed ratios of multivalents to bivalent pairs at metaphase I for each chromosome in each line (25 ratios) were tested for agreement with the theoretical 2:1 ratio (Table 4). It can be seen from the data in Table 4 and from Figure 2 that the autotetraploid lines included in this study frequently deviated from a 2:1 ratio, but in different directions. The C plants showed multivalent frequencies consistently in excess of 66.66% and three of the five chromosomes in this line (1, 3, and 5) showed significant excesses of multivalents when tested for agreement with a 2:1 ratio. In the E lines some significant excesses of multivalents also occurred, affecting chromosome 1 (all lines), chromosome 3 (lines E1, E2, and E3), and chromosome 5 (line E1). In contrast, the shorter chromosomes (2 and 4) consistently showed multivalent frequencies well below 66.66% and in most cases these deviated significantly from a 2:1 ratio (lower). In addition, chromosome 3 of line E4 also showed a significant (low) departure from 2:1.

View this table:

Semihierarchical (nested) analysis of variance of multivalent frequencies (after angular transformation) conducted on a reduced data set to test for between-plant variation within lines

Multivalent frequency and chiasma frequency: Chiasma frequency is one of the parameters that influences multivalent frequency at metaphase I (see discussion) and the only one that can be assessed directly by observations made at this stage. To examine the relationship of these parameters, mean multivalent frequency was plotted against mean chiasma frequency for each chromosome separately and also for the cell totals, and regression analyses were carried out for each situation. There is some suggestion of a positive relationship between chiasma frequency and multivalent frequency, particularly for chromosome 5, where a significant regression obtains, and for cell totals. In these, and one other comparison (chromosome 2), C plants have both the highest multivalent and chiasma frequencies whereas line E4 has both the lowest multivalent and chiasma frequencies. However, a statistically significant relationship between these parameters could not be consistently demonstrated. This could be due to the low number of lines (five) included in the study or to other factors obscuring an underlying relationship. An inspection of the data revealed several cases of differences in multivalent frequencies between pairs of lines whose chiasma frequencies were the same or very similar. These mostly concerned comparisons of line E4 with other lines and, as expected, affected chromosomes 2 and 4 but also chromosome 3.

View this table:

Analysis of variance of multivalent frequencies (after angular transformation) in four established (E) and one new (C) autotetraploid lines of Arabidopsis


The only previous cytogenetic analysis of meiosis in autotetraploid Arabidopsis involved a stable established line, generated by colchicine treatment of the Wilna accession, and subsequently maintained for 20-30 generations (Weiss and Maluszynska 2000). This line was reported to have 10 bivalents in ∼50% of the analyzed cells. The remaining cells had nine or eight “associations,” implying either one or two multivalents per cell, although the authors considered that some of these associations could result from overlapped bivalents. The authors conceded that this is a surprisingly low multivalent frequency (maximum 15% per tetrasome). They suggested that this could reflect the small size of Arabidopsis chromosomes, with effects on chromosome pairing/synapsis, but they also proposed that a process of diploidization might have occurred in this line during the 20-30 generations since its origin by chromosome doubling, possibly involving genome rearrangements. Limited evidence for the latter proposal was presented in the form of a translocation of a chromosome segment bearing 45S rDNA.

Figure 2.

—Multivalent frequencies (M) plotted against relative chromosome lengths for each of the five autotetraploid lines. Numbers along the top x-axis indicate the five different chromosomes. The boldface dashed horizontal line indicates the theoretical multivalent frequency on the random-end pairing model (66.6%) and the two lighter dashed horizontal lines give the multivalent frequency limits for agreement/departure from this theoretical frequency (5% probability); all points outside these limits depart significantly from the 2:1 ratio of multivalent:bivalent pairs.

The present study effectively disposes of the idea that Arabidopsis autotetraploids might be predisposed to low multivalent frequencies and high bivalent frequencies due to small chromosome size and/or some features of their pairing/synaptic behavior. The new (C) autotetraploids exhibited a fairly typical autotetraploid meiotic behavior with a high level of multivalent association at metaphase I (79% averaged over the five chromosomes). The frequency of multivalent synapsis at pachytene of prophase I is likely to be even higher, since not all pachytene multivalents necessarily persist to metaphase I, but in any event cannot be <79%. This level of metaphase I multivalent formation (197 multivalents:53 bivalent pairs) significantly exceeds the 2:1 ratio (66.66% multivalents) expected on the random-end pairing model [χ2(1) = 17.30; P < 0.001], implying that, despite their small size, Arabidopsis chromosomes have more than two autonomous pairing sites. A high frequency of multivalent pairing in Arabidopsis autotetraploids had already been predicted (Lavania 1991) on the basis of the observation of a high frequency of double reduction in a genetical analysis carried out by Van der Veen and Blankestijn-de Vries (1973).

These findings support the proposition (Jones and Vincent 1994) that interspecific multivalent frequency variation is not a simple function of chromosome size. For example, Crepis capillaris and T. monococcum have similarsized chromosomes, but Crepis autotetraploids have many more multivalents per tetrasome (86%) than Triticum (55%), while multivalent frequency in autotetraploid Arabidopsis (79%) approaches that seen in C. capillaris, despite having ∼10-fold smaller chromosomes. Factors affecting the organization of chromosome pairing/synapsis, other than simply chromosome size, must be involved in determining multivalent formation. Santos et al. (1995) have shown that in Secale cereale, despite its relatively large chromosomes, pairing partner choice in autotetraploids is restricted, in most cases, to two distally located APSs per tetrasome. As a result, these tetraploids rarely have more than one PPS per tetrasome at prophase I and multivalent frequencies at metaphase I are accordingly equal to or less than the 66% predicted by the random-end pairing model.

The four established autotetraploid lines included in this study formed significantly fewer multivalents than the two new autotetraploids did, but the extent of this reduction was strongly chromosome and line dependent. There is some uncertainty concerning the closeness of the E lines and C plants used in this study, which is almost unavoidable given the aim of comparing “historical” established tetraploid material with new colchicine-generated tetraploids. Nevertheless, we can be confident that all the material used belongs to the Col accession, as verified by microsatellite loci comparisons (see materials and methods) and the characteristic distribution of 5S and 45S rDNA chromosomal sites (Sanchez-Moranet al. 2002). Accepting this limitation, the results of this comparison strongly suggest that the autotetraploid E lines have undergone a process of partial diploidization of meiotic chromosome behavior during the 13 generations that they have been maintained. It would be interesting to test this hypothesis by examining meiotic behavior in naturally occurring autotetraploid materials, such as the Stockholm accession (Heslop-Harrison and Maluszynska 1994), which have presumably existed at the tetraploid level for much longer. However, this would involve comparisons between accessions that might indeed be genetically quite different (e.g., Sanchez-Moranet al. 2002). Another, longer-term approach could be to investigate meiotic behavior in successive generations after colchicine induction of autotetraploidy. In the meantime, work is proceeding to investigate additional colchicine-induced autotetraploids and preliminary findings confirm that they, too, have very high quadrivalent frequencies, comparable to those of the two C plants in this study (data not shown).

View this table:

χ 2 values testing goodness of fit to 2:1 ratio of multivalents: bivalent pairs for each chromosome (1-5) in each line

Diploidization of autopolyploids is thought to have occurred relatively frequently during the evolution of plant species, since many cases of apparently diploid plant genomes that have undergone ancestral genome duplication are now recognized (Mitchell-Olds and Clauss 2002). The mechanism(s) involved in diploidization, however, are less well known and more conjectural, involving the progressive accumulation of chromosome structural rearrangements accompanied by the introduction or strengthening of pairing/synapsis controls. Pairing/synapsis controls may also operate on preexisting genetic differences between identical vs. homologous chromosomes in autopolyploids (Santoset al. 1983). Alternatively, bivalent frequency may be promoted by reductions in the numbers of APSs; in the extreme case of just a single APS per chromosome, only bivalents will result. Similarly, bivalent associations may occur at metaphase I following multivalent synapsis due to chiasma insufficiency or an altered distribution of chiasmata relative to pairing partner exchanges. In these two latter situations, however, the plants will exhibit tetrasomic gene segregation ratios despite having exclusively bivalent synapsis, and so technically these cannot be regarded as diploidization in the genetical sense. Studies involving induced autopolyploids indicate that detectable amounts of meiotic diploidization may occur relatively rapidly with or without accompanying mutagenesis to induce structural chromosome rearrangements (Gillies 1989; Lavania 1991 and references therein). Other possible mechanisms leading to diploidization could include genetic effects such as duplicate gene silencing, inactivation, or divergence, as has been proposed for allopolyploids (Pickard 2001).

The reductions in multivalent frequencies and concomitant increases in bivalent frequencies shown by the E lines are strongly chromosome dependent. The likeliest cause of this is chromosome size-dependent change in pairing/synapsis of homologous chromosomes and/or in chiasma frequency/distribution. The shorter chromosomes of the genome may respond more readily to selection for alteration in these parameters. This would not be surprising since in general it is recognized that the incidences of PPSs, and hence of APSs, as well as chiasmata, decrease with decreasing chromosome size, and this size dependency may be enhanced in the E lines. However, a specific effect may be attributable to some special properties of chromosomes 2 and 4. These chromosomes are the smallest members of the genome and are structurally distinct from the other chromosomes, being acrocentric and possessing large heterochromatic NORs in their short arms. Chromosome 2, in particular, exhibits especially low multivalent frequencies in three of the E lines, even lower than chromosome 4 does. This indicates that some factor(s) other than size may be operating since chromosomes 2 and 4 are very similar in size and organization, at least at the level of gross morphology. Previous analyses of chiasma frequency variation in wild-type accessions and mutant lines of Arabidopsis have also indicated that chromosome 2 exhibits unusual properties, the causes of which are unknown (Sanchez-Moran et al. 2001, 2002).


We thank J. P. M. Camacho and F. Perfectii for comments on the manuscript. This work was supported by projects PB98-0719-CO2-02-No8320 and BMC2002-01171 awarded by Ministerio de Ciencia y Tecnologia (Spain).


  • Communicating editor: J. Birchler

  • Received March 10, 2003.
  • Accepted July 25, 2003.


View Abstract