Meiotic aberrations in the three genotypes
The basis of our analysis is the comparison of morphological and quantitative meiotic characteristics between three cauliflower genotypes that differ in the rates of offspring with aberrant phenotype. The so-called Low phenotype produces less than 5% aberrant offspring, the Moderate phenotype between 5 and 10% and the High phenotype more than 10%. For each genotype we studied around 50 DAPI stained pollen mother cells (PMCs), between pachytene – anaphase II, of which the majority were at diakinesis (more than 50 PMCs per genotype were observed).
Figures 1 (a, b) show representative examples of meiotic prophase cells at pachytene. Chromosomes at this stage in all genotypes are fully paired and display the typical pattern of brightly fluorescing pericentromeres, that are often tightly clumped forming a dense synizetic knot (Moens 1964), often considered the result of an acetic fixation artefact (Armstrong et al. 2001). The distal parts of the chromosome arms are the less condensed euchromatin with various smaller chromomeres. Although in some nuclei smaller chromosome regions are not paired, it is quite rare in our observation, we consider chromosome behaviour at this stage synaptic. Incidentally we distinguished in the three genotypes associations between non-homologous centromeres or pericentromeres (Figure 1a).
At diakinesis, chromosomes are highly condensed and chiasmata bonds, resulting from diplotene chiasmata are clearly distinguishable (Figure 2a-e). For the interpretation of the bivalent configurations, since chromosome arms can have only one chiasma, so two adjacent chiasmata in one arm will be interpreted as a single chiasma. Without at least one chiasma homologues are not attached and the two chromosomes remain as a univalent pair (Figure 2c). With only one chiasma between homologues, bivalents attain the shape of a cross or that of a rod, depending on the position of the chiasma (Figure 2a). With two chiasmata, a bivalent forms different shapes depending on proximal position, i.e., close to centromere or distal position, close to the chromosome end (Figure 2b). Most cells contained ring and rod bivalents and univalent pairs (Figure 2 d, e). In the Low genotype, univalents are rare and occur in only 10.5% of the cells. In the Moderate and High genotypes univalents are much more frequent (39% and 42.3% respectively). Interchromosomal connections between the homologous and non-homologous chromosomes are seen in all genotypes, and in some cells of the Moderate and High genotypes, almost all chromosomes exhibit these threads (as seen in Figure 2f).
At metaphase I the bivalents transgress to the equatorial plate with centromeres facing the poles. In a few cases, we observed multivalent-like configurations due to interchromosomal connections as mentioned above (Figure 2f). At anaphase I, the half bivalents disjoin normally, whereas the univalents segregate randomly resulting in unbalanced chromosome numbers at later stages. Moreover, anaphase bridges and multiple lagging chromosomes appeared (Figure 2g) in the Moderate and High genotypes. At anaphase II, we observed anaphase bridges and chromosome stickiness (interchromosomal connections) (Figures 2g, h). In cells at the tetrad stage, we found mostly balanced tetrads (9+9+9+9) in the Low genotype, while in the Moderate and High genotypes, tetrads with unbalanced chromosome numbers like (9+9, 9+8+1) and (10+8, 10+8) were observed (Figure 2i).
Number of chiasmata bonds in the three genotypes
Since the Moderate and High genotypes demonstrate relatively high numbers of non-chiasmatic chromosomes, we estimate the total number of chiasmata based on the numbers of counted rod and ring bivalents (with the assumption that they account for 1 and 2 chiasmata, respectively, Table 1, Figure 3b). Chiasma estimates for the Low, Moderate and High genotypes were 14, 13 and 14 per cell respectively, which is statistically not different (P = 0.152). Hence, we find no clear relation between the number of chiasmata bonds and the number of aneuploid offspring.
We therefore compared the mean number of estimated chiasmata in cells with univalent pairs with those that lack univalents. Our data suggests that the presence of univalent pairs leads to an increase of chiasmata on other chromosome pairs within the cell complement (Table 3).
More univalents in the Moderate and High genotypes
The frequencies of univalent pairs, rod bivalents, ring bivalents at diakinesis cells for the three genotypes are summarized in Table 1. The percentages of cells showing an expected number of nine bivalents are 89%, 61% and 57% for the Low, Moderate and High genotypes respectively. In most of the cells which contained univalents only one univalent pair was present, while in the Moderate and High genotypes, cell complements with two or more pairs of univalents were regularly observed. In the Low genotype, 7% of the cells have one pair of univalents and 4% of cells have two pairs of univalent. In the Moderate genotype, 27% of the cells have one pair of univalents and 12% of cells have two pairs of univalents. In the High genotype, 31% of the cells have one pair of univalents and 12% of cells have two or more pairs of univalents. To determine whether the three genotypes differ significantly in univalent incidence, a one-way ANOVA test on univalent counts showed significant differences between the genotypes (P = 0.002) (Figure 3a). In the Fisher's protected least significant difference test (series of pairwise t-tests) we showed that the Low genotype differs significantly from both the Moderate and High genotypes, while the Moderate and the High genotypes do not differ from one another (Table 2).
Immunodetection of chiasma precursor MLH1
We further investigated the nature of the chiasmata bonds by performing an immunofluorescence detection with antibodies against the chiasma precursor MLH1 on spread diakinesis cells. MLH1 is known to occur in the later recombination nodules of the synaptonemal complex at pachytene and is involved in the class I crossover pathway. The MLH1 foci in our slides clearly correspond to most of the chiasma sites (Figure 4) but are lacking in some places that may represent sites of class II crossovers or class I crossover sites that were not detected by the immune fluorescent method. We analysed at least 15 cells for each genotype showing on average 10.93, 10.43 and 9.52 MLH1 sites per cell in the Low, Moderate and High genotype, respectively. Compared to the number of chiasmata bonds, we calculated the percentage of MLH1 stained chiasmata at 77.5 % for the Low and the Moderate type, and 69 % for the High genotype (Table 1).
Interchromosomal connections
Besides univalent formation, the presence of bivalent connections is another remarkable feature of this cauliflower. We counted numbers of connections between bivalents (Table 1). The average numbers of connections were 1.1, 2.0 and 2.1 in the Low, Moderate, and High genotypes, respectively. A one-way ANOVA suggests these differences in bivalent interconnections between the three lines are significant (P-value < 0.001), and a subsequent Fisher's protected least significant difference test indicates that the Low genotype differs significantly from both the Moderate and High genotype (P < 0.001). However, the difference between the Moderate and High genotype is not significant (0.5286). Comparison of the three lines for the correlation of numbers of crossovers and numbers of interchromosomal connectives (Figure 3c) revealed that both variables were not significant (Kendall’s Tau coefficient: Low = 0.982, Moderate = 0.834, High = 0.375).
Repetitive DNA FISH analysis of diakinesis / metaphase I
To investigate if interchromosomal connections contain repetitive sequences, as was previously suggested by Pedrosa et al. (2001) in the study of Ornithogalum longibracteatum (Hyacinthaceae). Therefore, we performed a FISH experiment with repetitive DNA: 45s rDNA and two Brassica specific centromere repeats. The result indicated that 45s rDNA as well as the two Brassica centromere repeats painted interchromosomal connections (Figure 5a, c). Centromere and 45S rDNA specific sequences involved around 70% of bivalent connections, while less than 30% of the connections comprised non-stained chromatin. When 45s rDNA connections were observed, one or both chromosomes involved carry a 45s rDNA locus, whereas when centromere repeats are involved, these connect the centromere regions of two homologues (Figure 5b, d).
We then asked as to whether chromosomal connections are random, or whether specific chromosomes are more often involved in interconnections. To this end we had a closer look at the interconnections. Cauliflower has nine chromosome pairs, two of which have 45s rDNA loci. Connections between chromosomes with 45s rDNA (45S-45S), between a chromosome with and without a 45s rDNA signal (45S-non) and between chromosomes without 45s rDNA signals (non-non) were quantified for a subset of 19 painted interconnections (see Table 4). The theoretical chances of connections between chromosomes (while assuming no preferential connections being formed) are given by the following chances: P(45s-45s) = 2/9*1/8 =1/36, P(45S-non) = 2/9*7/8+7/9*2/8 = 14/36 and P(non-non) = 7/9*6/8 =21/36. A computed χ2 test of 27.8, and pdf=2< 0.001, clearly shows a significant overrepresentation of 45S rDNA repeats in interchromosomal connections (Table 5). At least for the 45s rDNA, the interconnections are not random, but preferentially occur between chromosomes that have functionally similar regions.
Meiotic analysis result of the APETALA1/ CAULIFLOWER Arabidopsis mutant
Since cauliflower is of all Brassica crops most plagued by aneuploidy in its offspring, we wondered whether the genes responsible for curd formation indirectly cause aneuploidy among offspring. We therefore studied the meiosis of the APETALA1/ CAULIFLOWER double mutant of Arabidopsis that, like Cauliflower, displays the typical curd phenotype (Figure 6a). Microscopic observations of pollen mother cells at diakinesis in this double mutant showed interchromosomal connections (Figure 6b, c), albeit far less than in Brassica: in less than 10% of the analysed cells. Univalents were not observed neither did these plants produce aneuploid offspring. We conclude that APETALA1 (AP1) and CAULIFLOWER (CAL) genes do not cause partial desynapsis during meiosis stages in Arabidopsis.